{"owner":"langbot-app","repo":"LangBot","hasSkills":true,"totalSkillsCount":156,"totalTokensCount":136227,"categories":["root-instruction","claude-rule","anthropic-skill"],"hasMcp":false,"mcpConfig":null,"found":["AGENTS.md","CLAUDE.md","skills/AGENTS.md","skills/README.md","skills/docs/user-guide.md","skills/package.json","skills/qa-agent-docs/qa-agent/00-technology-options.md","skills/qa-agent-docs/qa-agent/01-qa-agent-harness-plan.md","skills/qa-agent-docs/qa-agent/02-log-guard-plan.md","skills/qa-agent-docs/qa-agent/03-agent-browser-qa-principles.md","skills/qa-agent-docs/qa-agent/04-black-box-e2e-roadmap.md","skills/qa-agent-docs/qa-agent/README.md","skills/qa-agent-docs/user-guide.md","skills/schemas/README.md","skills/schemas/case.schema.json","skills/schemas/skill-index.schema.json","skills/schemas/suite.schema.json","skills/schemas/troubleshooting.schema.json","skills/skills.index.json","skills/skills/langbot-deploy/SKILL.md","skills/skills/langbot-dev/SKILL.md","skills/skills/langbot-eba-adapter-dev/SKILL.md","skills/skills/langbot-env-setup/SKILL.md","skills/skills/langbot-env-setup/references/browser-access-selection.md","skills/skills/langbot-env-setup/references/computer-use.md","skills/skills/langbot-env-setup/references/oauth-browser-profile.md","skills/skills/langbot-env-setup/references/playwright-mcp.md","skills/skills/langbot-env-setup/references/proxy.md","skills/skills/langbot-env-setup/references/service-startup.md","skills/skills/langbot-env-setup/references/wsl-notes.md","skills/skills/langbot-mcp-ops/SKILL.md","skills/skills/langbot-plugin-dev/SKILL.md","skills/skills/langbot-plugin-dev/references/test-env-setup.md","skills/skills/langbot-skills-maintenance/SKILL.md","skills/skills/langbot-skills-maintenance/references/curation-workflow.md","skills/skills/langbot-space-ops/SKILL.md","skills/skills/langbot-testing/SKILL.md","skills/skills/langbot-testing/cases/acp-agent-runner-debug-chat.yaml","skills/skills/langbot-testing/cases/agent-run-ledger-audit.yaml","skills/skills/langbot-testing/cases/agent-runner-async-db-readiness.yaml","skills/skills/langbot-testing/cases/agent-runner-behavior-matrix.yaml","skills/skills/langbot-testing/cases/agent-runner-fixture-contract.yaml","skills/skills/langbot-testing/cases/agent-runner-health-visibility.yaml","skills/skills/langbot-testing/cases/agent-runner-ledger-concurrency.yaml","skills/skills/langbot-testing/cases/agent-runner-ledger-contention.yaml","skills/skills/langbot-testing/cases/agent-runner-ledger-invariants.yaml","skills/skills/langbot-testing/cases/agent-runner-ledger-stress.yaml","skills/skills/langbot-testing/cases/agent-runner-live-install.yaml","skills/skills/langbot-testing/cases/agent-runner-qa-debug-chat.yaml","skills/skills/langbot-testing/cases/agent-runner-release-preflight.yaml","skills/skills/langbot-testing/cases/agent-runner-runtime-chaos.yaml","skills/skills/langbot-testing/cases/bot-event-routing-product-flow.yaml","skills/skills/langbot-testing/cases/box-mcp-heartbeat-recovery.yaml","skills/skills/langbot-testing/cases/claude-code-agent-debug-chat.yaml","skills/skills/langbot-testing/cases/codex-agent-debug-chat.yaml","skills/skills/langbot-testing/cases/dify-agent-debug-chat.yaml","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-cross-pipeline-isolation.yaml","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-fault-recovery.yaml","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-load.yaml","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-slow-load.yaml","skills/skills/langbot-testing/cases/langbot-fault-taxonomy-contract.yaml","skills/skills/langbot-testing/cases/langbot-live-backend-latency.yaml","skills/skills/langbot-testing/cases/langbot-live-backend-log-health.yaml","skills/skills/langbot-testing/cases/langbot-live-control-plane-api.yaml","skills/skills/langbot-testing/cases/langbot-overhead-accounting-contract.yaml","skills/skills/langbot-testing/cases/langbot-space-debug-chat-concurrency-smoke.yaml","skills/skills/langbot-testing/cases/langrag-kb-retrieve.yaml","skills/skills/langbot-testing/cases/langrag-parser-golden-e2e.yaml","skills/skills/langbot-testing/cases/langrag-sentinel-kb-discover.yaml","skills/skills/langbot-testing/cases/local-agent-basic-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-combo-rag-compaction-tool-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-complex-coding-task-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-context-compaction-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-effective-prompt-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-model-fallback-before-first-chunk-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-multimodal-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-multitool-rag-compaction-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-nonstreaming-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-parallel-tools-rag-compaction-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-plugin-tool-call-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-rag-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-rag-multimodal-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-steering-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-streaming-post-commit-failure-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-tool-error-recovery-debug-chat.yaml","skills/skills/langbot-testing/cases/local-agent-tool-loop-limit-debug-chat.yaml","skills/skills/langbot-testing/cases/mcp-stdio-register.yaml","skills/skills/langbot-testing/cases/mcp-stdio-tool-call.yaml","skills/skills/langbot-testing/cases/pipeline-debug-chat-performance.yaml","skills/skills/langbot-testing/cases/pipeline-debug-chat.yaml","skills/skills/langbot-testing/cases/plugin-e2e-smoke.yaml","skills/skills/langbot-testing/cases/provider-deepseek.yaml","skills/skills/langbot-testing/cases/qa-plugin-smoke-live-install.yaml","skills/skills/langbot-testing/cases/sandbox-skill-authoring-e2e.yaml","skills/skills/langbot-testing/cases/sandbox-skill-authoring-edit-existing-e2e.yaml","skills/skills/langbot-testing/cases/skill-discovery-via-mcp-gateway.yaml","skills/skills/langbot-testing/cases/webui-login-state.yaml","skills/skills/langbot-testing/cases/wizard-onebot-agent-runtime.yaml","skills/skills/langbot-testing/cases/wizard-runner-marketplace-catalog.yaml","skills/skills/langbot-testing/cases/wizard-scenario-routing.yaml","skills/skills/langbot-testing/cases/workspace-compatibility-preflight.yaml","skills/skills/langbot-testing/cases/workspace-plugin-pages-smoke.yaml","skills/skills/langbot-testing/cases/workspace-repository-contracts.yaml","skills/skills/langbot-testing/references/agent-runner-qa-workflow.md","skills/skills/langbot-testing/references/agent-runner-release-gate.md","skills/skills/langbot-testing/references/dify-agent-runner.md","skills/skills/langbot-testing/references/langrag-knowledge-base.md","skills/skills/langbot-testing/references/local-agent-runner-coverage.md","skills/skills/langbot-testing/references/local-agent-runner.md","skills/skills/langbot-testing/references/mcp-stdio-testing.md","skills/skills/langbot-testing/references/model-provider-testing.md","skills/skills/langbot-testing/references/performance-reliability-testing.md","skills/skills/langbot-testing/references/pipeline-debug-chat.md","skills/skills/langbot-testing/references/plugin-e2e-smoke.md","skills/skills/langbot-testing/references/sandbox-skill-authoring.md","skills/skills/langbot-testing/references/skill-all-tool-acceptance.md","skills/skills/langbot-testing/references/troubleshooting.md","skills/skills/langbot-testing/references/web-ui-testing.md","skills/skills/langbot-testing/references/workspace-release-testing.md","skills/skills/langbot-testing/suites/agent-runner-release-gate.yaml","skills/skills/langbot-testing/suites/core-smoke.yaml","skills/skills/langbot-testing/suites/eba-release-gate.yaml","skills/skills/langbot-testing/suites/langbot-debug-chat-isolation-gate.yaml","skills/skills/langbot-testing/suites/langbot-debug-chat-load-gate.yaml","skills/skills/langbot-testing/suites/langbot-live-backend-gate.yaml","skills/skills/langbot-testing/suites/langbot-performance-contract-gate.yaml","skills/skills/langbot-testing/suites/langbot-performance-reliability-gate.yaml","skills/skills/langbot-testing/suites/langbot-user-path-performance-gate.yaml","skills/skills/langbot-testing/suites/langbot-workspace-contract-gate.yaml","skills/skills/langbot-testing/suites/langbot-workspace-release-gate.yaml","skills/skills/langbot-testing/suites/local-agent-gate.yaml","skills/skills/langbot-testing/troubleshooting/agent-runner-actor-context-fields.yaml","skills/skills/langbot-testing/troubleshooting/aiosqlite-connect-hangs.yaml","skills/skills/langbot-testing/troubleshooting/ambiguous-runner-default-label.yaml","skills/skills/langbot-testing/troubleshooting/backend-not-listening.yaml","skills/skills/langbot-testing/troubleshooting/box-runtime-silent-action-timeout.yaml","skills/skills/langbot-testing/troubleshooting/box-session-conflict-logical-metadata.yaml","skills/skills/langbot-testing/troubleshooting/debug-chat-history-contaminates-automation.yaml","skills/skills/langbot-testing/troubleshooting/dynamic-form-missing-config-id.yaml","skills/skills/langbot-testing/troubleshooting/e2b-extra-mount-sync-missing.yaml","skills/skills/langbot-testing/troubleshooting/local-agent-mcp-text-result-empty-follow-up.yaml","skills/skills/langbot-testing/troubleshooting/local-agent-model-route-unavailable.yaml","skills/skills/langbot-testing/troubleshooting/marketplace-network-flaky.yaml","skills/skills/langbot-testing/troubleshooting/mcp-stdio-args-not-applied.yaml","skills/skills/langbot-testing/troubleshooting/nsjail-cli-compatibility.yaml","skills/skills/langbot-testing/troubleshooting/pipeline-form-controlled-warning.yaml","skills/skills/langbot-testing/troubleshooting/plugin-dependency-install-offline.yaml","skills/skills/langbot-testing/troubleshooting/plugin-runtime-timeout.yaml","skills/skills/langbot-testing/troubleshooting/provider-image-parse-error.yaml","skills/skills/langbot-testing/troubleshooting/proxy-env-mismatch.yaml","skills/skills/langbot-testing/troubleshooting/sandbox-native-tools-unavailable.yaml","skills/skills/langbot-testing/troubleshooting/socks-proxy-without-socksio.yaml","skills/skills/langbot-testing/troubleshooting/survey-widget-blocks-debug-chat.yaml","skills/skills/langbot-testing/troubleshooting/telemetry-proxy-noise.yaml","skills/skills/langbot-testing/troubleshooting/tool-name-collision-between-mcp-and-plugin.yaml","skills/skills/langbot-testing/troubleshooting/uv-run-resyncs-local-sdk.yaml"],"skills":{"AGENTS.md":"# AGENTS.md\n\nThis file guides code agents working in the LangBot main repository. `CLAUDE.md` is a symlink to this file.\n\nRead `ARCHITECTURE.md` before non-trivial backend, frontend, runtime, plugin, Box, MCP, persistence, or cross-repo SDK changes. This file is the working checklist; `ARCHITECTURE.md` is the system map.\n\n## Quick Facts\n\n- Python backend: `>=3.11,<4.0`, dependencies managed by `uv`.\n- Frontend: `web/` is Vite + React Router 7 + shadcn/ui + Tailwind, managed by `pnpm`.\n- Backend framework: Quart served by Hypercorn on `api.port`, default `5300`.\n- Frontend dev server: `web/` on `3000`, with `VITE_API_BASE_URL` pointing at the backend.\n- Plugin/Box/runtime contracts live in sibling repo `langbot-plugin-sdk`, pinned as `langbot-plugin` in `pyproject.toml`.\n\n## Essential Commands\n\n```bash\nuv sync --dev\nuv run main.py\nuv run pre-commit install\n\ncd web\npnpm install\npnpm dev\npnpm build\n```\n\nUseful focused tests:\n\n```bash\nuv run pytest tests/unit_tests -q\nuv run pytest tests/integration -q\nuv run pytest tests/integration/persistence -q\nuv run pytest tests/manual/mcp_smoke.py\n\ncd web\npnpm lint\npnpm test:e2e\n```\n\nRun the narrowest useful test first, then broader checks when confidence is needed.\n\n## Where to Look\n\n- Architecture map: `ARCHITECTURE.md`.\n- Dev environment guide: https://langbot.app/docs/zh/develop/dev-config.\n- Plugin runtime / CLI / SDK debugging: https://langbot.app/docs/zh/develop/plugin-runtime.\n- API-key auth: `docs/API_KEY_AUTH.md`.\n- Service API CLI: [`langbot-cli`](https://github.com/langbot-app/langbot-cli) (`lbctl`) manages running Workspaces; it is separate from the SDK's `lbp` CLI.\n- Box deep-dive notes: `docs/review/box-architecture.md` and related files.\n- In-repo skills: `skills/` is the single source of truth for LangBot agent skills.\n- SDK repo: `../langbot-plugin-sdk/` when changing shared entities, plugin APIs, action protocol, `lbp rt`, or `lbp box`.\n\n## Cross-Repo SDK Work\n\nWhen changing SDK contracts used by LangBot:\n\n```bash\n# from langbot-plugin-sdk, with LangBot's .venv active\nuv pip install .\n\n# from LangBot, preserve the locally installed SDK\nuv run --no-sync main.py\n```\n\nFor standalone runtime debugging:\n\n```bash\n# in langbot-plugin-sdk\nuv run --no-sync lbp rt\nuv run --no-sync lbp box\n\n# in LangBot\nuv run --no-sync main.py --standalone-runtime\nuv run --no-sync main.py --standalone-box\n```\n\nConfig keys to verify in `data/config.yaml` / `src/langbot/templates/config.yaml`:\n\n- Plugin runtime: `plugin.runtime_ws_url`, default Docker host `langbot_plugin_runtime:5400/control/ws`.\n- Box runtime: `box.enabled`, `box.backend`, `box.runtime.endpoint`, Docker host `langbot_box:5410`.\n- API/MCP auth: `api.global_api_key`.\n\n## Change Rules\n\n- HTTP API changes that should be agent-accessible must update the matching MCP tool in `src/langbot/pkg/api/mcp/server.py` and the relevant skill under `skills/` in the same pass.\n- When changing Service API routes or capabilities used by `lbctl`, check compatibility with the separate `langbot-cli` repository.\n- New schema changes use Alembic under `src/langbot/pkg/persistence/alembic/versions/`. LangBot 4.x does not support upgrading 3.x databases.\n- New platform behavior belongs in platform adapters only for platform translation; pipeline/business logic belongs in `pkg/pipeline/` or services.\n- User-facing strings must support i18n; follow the backend language rules below for server-generated content.\n- Conversation payload columns, including new ones, must use `MediaReferenceText` and persist image references only. See the [media persistence contract](ARCHITECTURE.md#persistence-and-configuration).\n- Interactive session lists use `SessionSummaryReader` to read only `monitoring_sessions`; `SessionSummaryWriter` maintains summaries in the source transaction. Historical initialization completes before writers start, never in requests or application background tasks. Follow the [conversation summary contract](ARCHITECTURE.md#conversation-summary-boundary) and keep `uv run pytest --required-suite session_summary` passing.\n- Code comments and docstrings must be English.\n- Keep compatibility and security in mind; LangBot is used in both self-hosted/community and toB deployments.\n- Commit message format: `(): `.\n\n## Backend Message Language\n\n- LangBot-owned business errors, runtime notices, and default messages sent to users must use the owning Workspace's message language. Keep this independent of each member's WebUI language; an unset Workspace language defaults to `en_US`.\n- Reuse `src/langbot/pkg/i18n/` and the existing `message_language` entry in `WorkspaceMetadata`. Use `t()` with stable English templates and named parameters; add matching keys and placeholders to all eight catalogs: `en_US`, `zh_Hans`, `zh_Hant`, `ja_JP`, `es_ES`, `ru_RU`, `th_TH`, and `vi_VN`. Avoid new translation frameworks or duplicate language settings.\n- Resolve language from a trusted Workspace or bot execution context, and isolate it across concurrent requests, tasks, and adapter callbacks. Browser language hints may supply account-only or unauthenticated response language, but must not override a resolved Workspace or authorize tenant selection. Reading/changing the setting requires `workspace.view` / `workspace.update`.\n- Pipeline notices should be configurable in Pipeline output settings through `pipeline_notice()`. Blank overrides inherit the Workspace translation. Preserve user-authored text literally except for documented placeholders, and retain existing visibility switches. Do not rewrite historical messages or custom text when the language changes.\n- China-only adapters and configuration/field validation diagnostics do not require localization. An adapter also serving international users is not exempt merely because it supports a Chinese platform.\n- Preserve raw provider/plugin diagnostics, tracebacks, protocol values, and machine-readable error/status codes. Translate LangBot's own category or explanation around them; do not translate arbitrary errors with substring replacement. Plugin-authored messages remain the plugin's localization responsibility.\n- Keep changes small and maintainable. Reuse existing persistence and response boundaries; delete obsolete fields only after verifying they have no remaining readers, including compatibility paths.\n- When changing language behavior, verify catalog/placeholder parity, Workspace isolation and permissions, and Pipeline inheritance/custom overrides. Update affected configuration snapshots and API/MCP documentation. Implementation details and focused test locations: [Backend message language](docs/architecture/backend-i18n.md).\n\n## Runtime Pitfalls\n\n- Local stdio Plugin Runtime disconnects do not auto-reconnect; restart LangBot if that path breaks.\n- Orphan runtime processes on `5400`/`5401` commonly break plugin debugging.\n- Use `uv run --no-sync` after locally installing the SDK, or `uv` may restore the pinned package.\n- A false Box “no backend” often means Docker is running but the current user lacks Docker socket permission.\n- Do not confuse external MCP servers LangBot connects to (`pkg/provider/tools/loaders/mcp.py`) with LangBot's own `/mcp` server (`pkg/api/mcp/`).\n- `CLAUDE.md` is a symlink to this file; edit `AGENTS.md`, not the symlink.\n\n## Principles\n\n- Keep it simple, stupid.\n- Entities should not be multiplied unnecessarily.\n- 八荣八耻\n\n 以瞎猜接口为耻，以认真查询为荣。\n 以模糊执行为耻，以寻求确认为荣。\n 以臆想业务为耻，以人类确认为荣。\n 以创造接口为耻，以复用现有为荣。\n 以跳过验证为耻，以主动测试为荣。\n 以破坏架构为耻，以遵循规范为荣。\n 以假装理解为耻，以诚实无知为荣。\n 以盲目修改为耻，以谨慎重构为荣。\n","CLAUDE.md":"AGENTS.md","skills/AGENTS.md":"# Agent Workflow\n\nThis repository stores reusable LangBot agent-testing assets. Keep changes structured so the next agent does not need to rediscover paths.\n\n## First Steps\n\n1. Read `skills/.env` before using local URLs, paths, browser profiles, or proxy defaults. If present, `skills/.env.local` overrides it for this machine and must not be committed. On a new machine, copy `skills/.env.example` to `skills/.env.local` first.\n2. Pick the smallest relevant skill:\n - `langbot-env-setup` for environment, browser, OAuth, proxy, and startup.\n - `langbot-testing` for WebUI, provider, pipeline, cases, and troubleshooting.\n - `langbot-skills-maintenance` for adding, deduplicating, or auditing this skills repository.\n3. Prefer existing cases and troubleshooting entries before exploring from scratch.\n\n## Editing Rules\n\n- UI/browser testing is the primary QA path. API/curl checks are diagnostic only and cannot make a UI case pass by themselves.\n- Put skills under `skills//`.\n- Keep `SKILL.md` concise; move detailed workflows to `references/`.\n- Put reusable test paths in `cases/*.yaml`.\n- New or edited cases must include `priority`, `risk`, `ci_eligible`, and `evidence_required` so agents can select the right test set without rereading every file.\n- Use `env_any` / `automation_env_any` for one-of machine inputs, such as `LANGBOT_PIPELINE_URL|LANGBOT_PIPELINE_NAME`; do not list those alternatives as separate all-required env keys.\n- Put reusable groups of cases in `suites/*.yaml` rather than hardcoding test sets in docs or CLI code.\n- Put growing failure knowledge in `troubleshooting/*.yaml`.\n- Do not hardcode local ports in testing docs; use `skills/.env` variables and machine-local `skills/.env.local` overrides.\n- Do not store secrets, API keys, OAuth tokens, or localStorage token values.\n\n## Required Checks\n\nAfter structural changes, run:\n\n```bash\nbin/lbs validate\n```\n\nAfter changing skills, cases, or troubleshooting assets, run:\n\n```bash\nbin/lbs index\n```\n\nUse `bin/lbs env show` to inspect defaults and `bin/lbs env doctor` when diagnosing local environment readiness. Env output is redacted by default; do not work around that by printing raw secrets.\n`bin/lbs` is a generated local wrapper. If it is missing on a fresh checkout, run `npm run bootstrap` from this directory first; `npm install` also regenerates it via `prepare`.\nUse `bin/lbs fixture check` before fixture-heavy cases such as MCP, RAG, multimodal, or plugin smoke tests.\nUse `bin/lbs case list --ready` for cases that have no missing machine inputs and no manual preconditions. Use `bin/lbs case list --machine-ready` when you want to keep `manual-check` candidates and confirm their preconditions yourself.\n\nBefore executing a saved QA path, generate the agent-facing plan:\n\n```bash\nbin/lbs test plan \n```\n\nRead the plan readiness sections before running the browser path. Missing env,\nautomation env, or fixture readiness means the case is not ready to execute and\nshould be marked `blocked` or fixed first.\n`manual_check` means machine inputs are present but the agent must verify the\ndeclared `preconditions` or `setup` items before executing the UI path. Do not\nturn a `manual_check` case into `pass` until those items were checked in the\nsame run.\n\nBefore executing a group of saved QA paths, generate the suite plan:\n\n```bash\nbin/lbs suite plan \n```\n\nUse `bin/lbs suite start ` to create a shared suite run id, suite evidence root, per-case evidence directories, and `suite-start.json`/`suite-start.md` handoff files. Then run `bin/lbs suite report --evidence-dir ` to aggregate case results.\nAutomation scripts write `automation-result.json`; write the final per-case `result.json` with `bin/lbs test result --result --reason --evidence-dir --evidence ` after collecting the required evidence. A `pass` result must include all required evidence.\nFor runner-specific Debug Chat cases, prefer case-specific pipeline env keys such as `LANGBOT_LOCAL_AGENT_PIPELINE_URL` over the generic `LANGBOT_PIPELINE_URL`; otherwise an agent can accidentally test the wrong runner.\n","skills/README.md":"# LangBot Skills\n\nThis directory is the **single source of truth** for LangBot's agent skills —\nreusable, on-demand instruction packs for AI agents (Claude Code, Codex, Cursor,\nand LangBot's own Local Agent) working with the LangBot ecosystem.\n\n> These skills were consolidated here from the former `langbot-app/langbot-skills`\n> repository (now archived). Documentation and the landing page link here; do not\n> re-copy skill content elsewhere — link to this directory instead.\n\n## Skill catalog\n\n| Skill | What it covers |\n| --- | --- |\n| [`langbot-dev`](skills/langbot-dev) | Core backend + web frontend development (Quart, Vite, API, migrations, MCP server) |\n| [`langbot-plugin-dev`](skills/langbot-plugin-dev) | Plugin SDK / component development, debugging, WebSocket testing |\n| [`langbot-deploy`](skills/langbot-deploy) | Docker / Compose / Kubernetes deployment, config.yaml, Box runtime, global API key |\n| [`langbot-testing`](skills/langbot-testing) | WebUI / e2e QA harness, cases, fixtures, troubleshooting (the `bin/lbs` CLI) |\n| [`langbot-env-setup`](skills/langbot-env-setup) | Local dev/test environment, browser access, OAuth, proxy, startup |\n| [`langbot-mcp-ops`](skills/langbot-mcp-ops) | Operating a LangBot instance through its MCP server (`/mcp`) |\n| [`langbot-space-ops`](skills/langbot-space-ops) | Browsing the LangBot Space marketplaces through the Space MCP server |\n| [`langbot-eba-adapter-dev`](skills/langbot-eba-adapter-dev) | Building platform adapters for the Event-Based Agents architecture |\n| [`langbot-skills-maintenance`](skills/langbot-skills-maintenance) | Adding, deduplicating, and auditing skills in this directory |\n\n`skills.index.json` is the machine-readable index (regenerate with `bin/lbs index`).\n\n## Quick start (for an AI agent)\n\n1. Read this README, `AGENTS.md`, and `docs/user-guide.md` to understand the layout.\n2. Read `skills/.env` for shared local defaults. On a new machine, copy\n `skills/.env.example` to `skills/.env.local` (gitignored) and override\n machine-specific values there. Never commit secrets.\n3. Pick the smallest relevant skill from the catalog above and follow its\n `SKILL.md`.\n\n## The `lbs` CLI\n\nThe testing assets ship with a small CLI (`bin/lbs`, Node >= 22.6). The\n`bin/lbs` wrapper is a generated local entrypoint; on a fresh checkout, run\n`npm run bootstrap` once if it is missing. `npm install` also regenerates it via\nthe `prepare` script.\n\n```bash\nnpm run bootstrap # create bin/lbs if missing\nbin/lbs validate # validate skills/cases/troubleshooting structure\nbin/lbs index # regenerate skills.index.json\nbin/lbs env show # inspect resolved env defaults (redacted)\nbin/lbs env doctor # diagnose local environment readiness\nbin/lbs case list --ready\nbin/lbs test plan \nbin/lbs suite plan langbot-debug-chat-load-gate\n```\n\n## Maintenance rule\n\nWhen the LangBot / LangBot Space **API or MCP server changes**, the\ncorresponding skill here MUST be updated in the same change. The MCP tool\nsurface, the API, and these skills are kept in lockstep — see each repo's\n`AGENTS.md`.\n","skills/docs/user-guide.md":"# LangBot QA Skills User Guide\n\nUse this guide as the first operational path after reading `README.md` and\n`AGENTS.md`.\n\n## 1. Configure Local Inputs\n\nRead `skills/.env`, then create `skills/.env.local` for machine-local values.\nDo not commit `.env.local`, browser profiles, reports, tokens, API keys, OAuth\nstate, or provider credentials.\n\nMinimum local fields for live browser QA:\n\n```bash\nLANGBOT_REPO=/path/to/LangBot\nLANGBOT_WEB_REPO=/path/to/LangBot/web\nLANGBOT_BACKEND_URL=http://127.0.0.1:5300\nLANGBOT_FRONTEND_URL=http://127.0.0.1:3000\nLANGBOT_DEV_FRONTEND_URL=http://127.0.0.1:3000\nLANGBOT_BROWSER_PROFILE=/path/to/langbot-browser-profile\nLANGBOT_CHROMIUM_EXECUTABLE=/path/to/chromium-or-playwright-chrome\nLANGBOT_E2E_LOGIN_USER=qa-local@example.com\n```\n\n`LANGBOT_E2E_LOGIN_USER` is a local QA account. The setup automation uses the\nLangBot recovery key from the active checkout to initialize or refresh that\nlocal account and write a browser `localStorage` token. It does not need the\nuser's GitHub or Space credentials.\n\n## 2. Check Readiness\n\nFrom `skills/`:\n\n```bash\nbin/lbs env show\nbin/lbs env doctor\nbin/lbs validate\nbin/lbs index --check\n```\n\n`env doctor` should report reachable backend and frontend URLs before live\nbrowser cases are run. Missing Space provider credentials are not a LangBot\nproduct pass; classify them as `env_issue` and configure the local Space\nprovider before measuring Debug Chat performance.\n\n## 3. Start Services\n\nStart the backend from `LANGBOT_REPO`:\n\n```bash\ncd \"$LANGBOT_REPO\"\nuv run main.py\n```\n\nStart the standalone frontend from `LANGBOT_WEB_REPO` and point it at the\nbackend:\n\n```bash\ncd \"$LANGBOT_WEB_REPO\"\nVITE_API_BASE_URL=\"$LANGBOT_BACKEND_URL\" pnpm dev --host 0.0.0.0\n```\n\nIf `VITE_API_BASE_URL` is missing, browser tests can load the Vite page but send\nAPI requests to the frontend port, which produces false UI failures.\n\n## 4. Prepare User-Path Fixtures\n\nFor local-agent Debug Chat cases and the user-path performance gate:\n\n```bash\nnode scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\n```\n\nThe script:\n\n- refreshes the local QA login and browser token;\n- marks the local wizard as skipped;\n- creates or updates a local QA pipeline;\n- scans Space LLM models, tests candidates, and switches to the first working\n Space model with tested fallback models;\n- writes `LANGBOT_PIPELINE_URL`, `LANGBOT_PIPELINE_NAME`, and local-agent\n pipeline/model variables into `skills/.env.local`;\n- returns `env_issue` when no Space model can be scanned or tested.\n\nUseful model controls:\n\n```bash\nLANGBOT_E2E_MODEL_TEST_LIMIT=8\nLANGBOT_E2E_MODEL_FALLBACK_COUNT=3\nLANGBOT_E2E_SKIP_MODEL_UUIDS=uuid-a,uuid-b\nLANGBOT_E2E_SKIP_MODEL_NAMES=model-a,model-b\nLANGBOT_E2E_SCAN_SPACE_MODELS=true\n```\n\nThe setup writes a current-runtime compatibility `max-round` value into the\npipeline config because this backend still reads that field directly during\nmessage truncation. Do not treat it as a long-term QA contract.\n\n## 5. Run Gates\n\nCross-repository contract gate (no model calls):\n\n```bash\nbin/lbs suite plan langbot-workspace-contract-gate\nbin/lbs suite run langbot-workspace-contract-gate --run-id langbot-workspace-contract-local\n```\n\nTop-down workspace release gate (browser workflows plus one complex Agent task):\n\n```bash\nbin/lbs suite plan langbot-workspace-release-gate\nbin/lbs suite run langbot-workspace-release-gate --run-id langbot-workspace-release-local --include-manual-check\n```\n\nRun `agent-run-ledger-audit` immediately after the complex Agent case. Set\n`LANGBOT_AGENT_RUN_ID` when the test instance has concurrent operator traffic.\n\nFast contract gate, no live service required:\n\n```bash\nbin/lbs suite run langbot-performance-contract-gate --run-id langbot-contract-local\n```\n\nLive backend gate:\n\n```bash\nbin/lbs suite run langbot-live-backend-gate --run-id langbot-backend-local\n```\n\nBrowser-visible user-path performance gate:\n\n```bash\nbin/lbs suite plan langbot-user-path-performance-gate\nbin/lbs suite run langbot-user-path-performance-gate --run-id langbot-user-path-local --include-manual-check\n```\n\nControlled Debug Chat message-path load gate (manual/non-required; run fake-provider cases serially when they share `LANGBOT_FAKE_PROVIDER_URL`):\n\n```bash\nbin/lbs suite plan langbot-debug-chat-load-gate\nbin/lbs test run langbot-fake-provider-debug-chat-load --run-id langbot-fake-load-local\nbin/lbs test run langbot-fake-provider-debug-chat-slow-load --run-id langbot-fake-slow-local\nbin/lbs test run langbot-fake-provider-debug-chat-fault-recovery --run-id langbot-fake-fault-local\nbin/lbs test run langbot-space-debug-chat-concurrency-smoke --run-id langbot-space-smoke-local\n```\n\nCross-pipeline Debug Chat isolation is a separate manual regression gate because\ncurrent releases may fail it due to product bug #2286:\n\n```bash\nbin/lbs suite plan langbot-debug-chat-isolation-gate\nbin/lbs suite run langbot-debug-chat-isolation-gate --run-id langbot-debug-chat-isolation-local --include-manual-check\n```\n\nStart with `langbot-fake-provider-debug-chat-load`. It launches a local\nOpenAI-compatible fake provider, creates the matching provider/model/pipeline,\nthen sends concurrent WebSocket Debug Chat messages through the real backend.\nUse `langbot-fake-provider-debug-chat-slow-load` to measure the same path under\ndeterministic streaming latency. Use\n`langbot-fake-provider-debug-chat-fault-recovery` to inject bounded provider\nHTTP failures and confirm later Debug Chat requests recover. Use the separate\n`langbot-debug-chat-isolation-gate` to verify that concurrent Debug Chat traffic\non two pipelines does not leak assistant responses across pipeline boundaries;\ncurrent releases may fail that gate because of #2286, so keep it out of the\nnormal load gate until the product fix lands.\nUse `langbot-space-debug-chat-concurrency-smoke` only as a low-volume live\nprovider smoke; it includes Space/model/network latency and should be compared\nagainst the fake-provider baseline before attributing failures to LangBot.\n\n`manual_check` means the agent must confirm the declared preconditions for that\nrun window. When setup automation is declared, run output may stop early with\n`env_issue`; fix that environment input before treating the product path as\nmeasured.\n\n## 6. Read Results\n\nSuite reports live under `skills/reports/`. Evidence lives under\n`skills/reports/evidence//`.\n\nFor performance cases, inspect:\n\n- `metrics.json` for p50/p95/p99, error rate, and total duration;\n- `automation-result.json` for threshold decisions and artifacts;\n- `console.log` and `network.log` for frontend/API failures;\n- backend logs for provider, runner, WebSocket, or persistence failures.\n\nDo not call a user-path performance result a LangBot overhead regression until\nprovider/tool/network time has been separated or ruled out.\n","skills/package.json":"{\n \"private\": true,\n \"type\": \"module\",\n \"bin\": {\n \"lbs\": \"./bin/lbs\"\n },\n \"scripts\": {\n \"bootstrap\": \"node scripts/bootstrap-lbs.mjs\",\n \"prepare\": \"node scripts/bootstrap-lbs.mjs\",\n \"prevalidate\": \"node scripts/bootstrap-lbs.mjs\",\n \"preindex\": \"node scripts/bootstrap-lbs.mjs\",\n \"preindex:check\": \"node scripts/bootstrap-lbs.mjs\",\n \"pretest\": \"node scripts/bootstrap-lbs.mjs\",\n \"precheck\": \"node scripts/bootstrap-lbs.mjs\",\n \"lbs\": \"node src/lbs.ts\",\n \"test\": \"node test/lbs-cli.test.ts && python3 -m unittest discover -s test -p 'test_*.py'\",\n \"validate\": \"node src/lbs.ts validate\",\n \"index\": \"node src/lbs.ts index\",\n \"index:check\": \"node src/lbs.ts index --check\",\n \"check:syntax\": \"find src test scripts -type f \\\\( -name '*.ts' -o -name '*.mjs' \\\\) -print0 | xargs -0 -n1 node --check\",\n \"check\": \"npm run check:syntax && npm run validate && npm test\"\n },\n \"engines\": {\n \"node\": \">=22.6\"\n },\n \"devDependencies\": {\n \"playwright\": \"^1.60.0\"\n }\n}\n","skills/qa-agent-docs/qa-agent/00-technology-options.md":"# LangBot Agent Testing 技术选型\n\n## 状态\n\n这是技术选型背景文档，不是当前路线图。当前黑盒 E2E QA 的实施顺序见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n## 目标\n\n`langbot-skills` 的目标不是替代测试框架，而是沉淀 agent 可复用的测试资产，让开发者 clone 仓库后，可以让 Codex、Claude Code、Computer Use 或 Playwright MCP 复用已有路径完成 LangBot 功能验证。\n\n核心原则：\n\n- Skill 负责路由和少量规则。\n- Reference 负责可读流程和背景知识。\n- Case 负责结构化测试路径。\n- Troubleshooting 负责结构化故障资产。\n- `lbs` 负责结构校验、索引、资产创建和未来的运行/报告能力。\n- UI/browser 是产品 QA 的主路径；API/curl 只用于诊断。\n\n## 浏览器控制层\n\n不同开发者可用的浏览器控制能力不同，所以浏览器层必须可替换。\n\n| 方案 | 适用场景 | 优点 | 代价 |\n|---|---|---|---|\n| Codex / Claude Computer Use | agent 可以直接控制可见浏览器 | 登录和交互路径最自然，通常不需要额外 MCP 浏览器桥接 | 依赖具体 agent 工具能力 |\n| Playwright MCP | 没有 Computer Use，但有 MCP 浏览器工具 | 稳定、可脚本化、适合回归路径 | OAuth 登录通常需要额外 visible profile |\n| 直接 Playwright 脚本 | 测试路径非常稳定，适合 CI | 可重复性强 | 需要维护脚本和 selector |\n| 商业 AI QA 平台 | 团队希望外包测试运行平台 | 报告和 PR 集成完整 | 成本和平台绑定 |\n\n## 当前推荐\n\n先采用分层降级：\n\n```text\n有 Computer Use？\n 是 -> 使用 Computer Use 控制浏览器\n 否 -> 使用 Playwright MCP\n\n需要 GitHub OAuth？\n 是 -> 使用持久浏览器 profile，让用户手动完成登录\n 否 -> 直接使用已有登录态或测试账号状态\n```\n\n具体选择逻辑沉淀在：\n\n```text\nskills/langbot-env-setup/references/browser-access-selection.md\n```\n\n测试原则固定在：\n\n```text\ndocs/qa-agent/03-agent-browser-qa-principles.md\n```\n\n## 环境变量层\n\n测试文档不应写死端口。共享默认值放在：\n\n```text\nskills/.env\n```\n\n关键变量：\n\n```text\nLANGBOT_FRONTEND_URL\nLANGBOT_BACKEND_URL\nLANGBOT_DEV_FRONTEND_URL\nLANGBOT_REPO\nLANGBOT_WEB_REPO\nLANGBOT_BROWSER_PROFILE\n```\n\nAgent 执行测试前应先读取 `skills/.env`，再用用户提供的当前环境或已启动服务覆盖默认值。\n\n## 测试资产层\n\n测试资产分两类：\n\n```text\nskills//\n references/ # Markdown 流程说明\n cases/ # 结构化测试用例\n troubleshooting/ # 结构化故障记录\n```\n\n当前已实现：\n\n- `SKILL.md` 路由\n- `references/*.md`\n- `lbs case new/list/show`\n- `lbs trouble show/search`\n- `lbs test plan`\n- `lbs test report`\n- `lbs list / validate / index`\n\n下一步重点：\n\n- 日志守卫规则补充\n- 报告产物管理\n\n## 关键判断\n\n不要强制所有内容只能通过 CLI 修改。更好的模式是：\n\n- 新增 case/troubleshooting：优先使用 `lbs`\n- 大段流程说明：允许直接编辑 Markdown\n- 结构性变更后：必须运行 `lbs validate`\n- 任何生成索引的变更后：运行 `lbs index`\n\n这样既能沉淀结构化资产，又不会在 schema 未稳定时拖慢迭代。\n","skills/qa-agent-docs/qa-agent/01-qa-agent-harness-plan.md":"# LangBot Skills 测试资产库规划\n\n## 状态\n\n这是早期测试资产库规划文档，保留用于解释 `langbot-skills` 的分层来源。\n\n当前路线已经收敛为黑盒 E2E QA：开发者用 agent 通过浏览器测试 LangBot，\n稳定路径沉淀为 case，失败知识沉淀为 troubleshooting。`lbs test report` 和\n日志守卫已有 MVP，后续重点是报告证据、case 元数据和少量稳定路径自动化。当前优先级见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n本文中关于 `case list/show`、`trouble show/search`、`test plan` 的“计划实现”\n内容已经部分过时，因为这些能力已经落地。\n\n## 目标\n\n让开发者 clone `langbot-skills` 后，可以把测试意图交给 agent，由 agent 复用已有环境配置、测试路径和故障知识完成 LangBot 功能验证。\n\n典型场景：\n\n- 冒烟测试：验证 pipeline Debug Chat、provider、常见页面是否正常。\n- Provider 测试：添加 DeepSeek/OpenAI/Claude 等供应商并验证模型可用。\n- 新 feature 测试：探索新 UI 路径，并在稳定后沉淀成 case/reference。\n- 回归测试：复用旧路径，避免每个窗口重新探索登录、模型配置、pipeline 调试。\n- 故障沉淀：把 runtime 超时、代理不一致、WebSocket 问题记录为可搜索资产。\n\n核心方向见 `03-agent-browser-qa-principles.md`：agent 必须以浏览器/UI 为主路径，API/curl 只能作为诊断手段。\n\n## 当前仓库结构\n\n```text\nskills/\n .env # 共享默认变量\n langbot-env-setup/ # 环境准备、浏览器控制路径、代理、登录态\n langbot-testing/ # WebUI / provider / pipeline 测试入口\n langbot-plugin-dev/ # 插件开发测试\n langbot-eba-adapter-dev/ # 平台适配器开发测试\nsrc/\n lbs.ts # CLI 源码\nbin/\n lbs # CLI 入口\ndocs/\n qa-agent/ # 规划文档，历史目录名保留\n```\n\n## 设计分层\n\n### 1. Skill 层\n\n`SKILL.md` 只做触发和路由，不承载大段流程。\n\n例子：\n\n```text\nlangbot-env-setup -> 选择 Computer Use / Playwright MCP / OAuth profile / proxy\nlangbot-testing -> 选择 WebUI / pipeline / provider / troubleshooting\n```\n\n### 2. Reference 层\n\nMarkdown 记录人和 agent 都能读的流程说明。\n\n适合内容：\n\n- 如何选择浏览器控制方式\n- 如何启动/检查服务\n- 如何执行 pipeline Debug Chat\n- 如何处理 OAuth 登录态\n\n### 3. Case 层\n\n使用 YAML 记录可重复测试路径。\n\n建议结构：\n\n```text\nskills/langbot-testing/cases/\n pipeline-debug-chat.yaml\n provider-deepseek.yaml\n```\n\n建议格式：\n\n```yaml\nid: pipeline-debug-chat\ntitle: Pipeline Debug Chat returns a bot response\nmode: agent-browser\narea: pipeline\ntype: smoke\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nsteps:\n - Open LANGBOT_FRONTEND_URL\n - Navigate to Pipelines\n - Open target pipeline\n - Select Debug Chat\n - Send deterministic prompt\nchecks:\n - \"UI: User message appears\"\n - \"UI: Bot message appears\"\n - \"Console: No unexpected frontend errors\"\n - \"Logs: Backend log includes Conversation(0) Streaming completed\"\ndiagnostics:\n - \"Use API/curl only after the UI path is attempted, to distinguish frontend display failure from backend/runtime failure.\"\ntroubleshooting:\n - plugin-runtime-timeout\n - proxy-env-mismatch\n```\n\n### 4. Troubleshooting 层\n\n故障资产会逐渐变大，适合结构化记录。\n\n历史 Markdown 入口保留在：\n\n```text\nskills/langbot-testing/references/troubleshooting.md\n```\n\n当前 canonical 结构化故障资产在：\n\n```text\nskills/langbot-testing/troubleshooting/\n plugin-runtime-timeout.yaml\n proxy-env-mismatch.yaml\n```\n\n### 5. CLI 层\n\n`lbs` 是统一入口，不再引入独立 `qa` 命令。\n\n已实现或当前可用：\n\n```bash\nbin/lbs list\nbin/lbs validate\nbin/lbs index\nbin/lbs new-skill \nbin/lbs new-ref \nbin/lbs case new pipeline-debug-chat --title \"Pipeline Debug Chat\"\nbin/lbs case list\nbin/lbs case show pipeline-debug-chat\nbin/lbs trouble list \nbin/lbs trouble show plugin-runtime-timeout\nbin/lbs trouble search runtime\nbin/lbs trouble add --title ... --symptom ... --cause ... --fix ...\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log\n```\n\n## 测试库位置\n\n不要使用隐藏 `.qa/` 作为主测试库。测试资产应该和 skill 放在一起，便于触发和维护：\n\n```text\nskills/langbot-testing/\n references/\n cases/\n troubleshooting/\n reports/ # 可选，本地运行产物可按需忽略或输出到外部目录\n```\n\n如果未来需要项目本地测试库，可以允许 `lbs` 支持 `--workspace` 或项目根目录配置，但 canonical 资产仍保存在 `langbot-skills`。\n\n## 阶段规划\n\n### 阶段一：环境和测试路径沉淀\n\n状态：基本完成，持续维护。\n\n- `skills/.env` 管共享默认变量。\n- `langbot-env-setup` 拆出 Computer Use、Playwright MCP、OAuth profile、proxy、service startup。\n- `langbot-testing` 记录 WebUI、pipeline、provider 测试路径。\n- `lbs validate/index` 维护结构。\n\n完成标准：\n\n- agent 可以从 `skills/.env` 和 references 中找到当前测试入口。\n- pipeline Debug Chat 这类路径不再需要从头探索。\n\n### 阶段二：结构化 case/troubleshooting\n\n状态：主体已完成，继续补齐元数据和资产质量。\n\n目标：\n\n- `lbs case new/list/show`\n- `lbs trouble show/search`\n- case id 去重、字段校验、索引生成\n\n完成标准：\n\n- 冒烟测试路径可以用结构化 case 表示。\n- 下一个 agent 窗口可以直接读取 case 执行。\n\n### 阶段三：计划和报告\n\n状态：已有 MVP，继续完善。\n\n目标：\n\n- `lbs test plan `\n- agent 按 plan 使用浏览器执行 UI QA\n- `lbs test report`\n- 日志守卫集成\n- 报告产物和 evidence 约定\n\n完成标准：\n\n- agent 可以按 case plan 执行浏览器测试。\n- 结果报告包含 UI 结果、后端日志、console 错误和 troubleshooting 建议。\n\n## 执行规则\n\n- agent 可以直接编辑 Markdown reference。\n- 新增结构化 case/troubleshooting 时，优先使用 `lbs`。\n- 每次结构变更后运行 `bin/lbs validate`。\n- 每次索引相关变更后运行 `bin/lbs index`。\n- 测试文档不写死端口，使用 `skills/.env` 中的 URL 变量。\n- 测试 case 的 `mode` 固定为 `agent-browser`。\n- API/curl 只能写入 `diagnostics`，不能替代 UI 步骤和 UI 检查。\n","skills/qa-agent-docs/qa-agent/02-log-guard-plan.md":"# 日志守卫规划\n\n## 状态\n\n这是当前活跃设计，已有第一版文件扫描 MVP。实现边界需要和黑盒 E2E 路线保持一致：\n\n- 日志守卫服务于 `lbs test report`。\n- 它不替代浏览器/UI 判断。\n- 它不发展成独立后端 API 测试框架。\n- 第一版默认扫描 `LANGBOT_REPO/data/logs/` 下最新的 `langbot-*.log`，也可扫描 agent\n 显式提供的 backend/frontend/console 日志文件。\n\n当前总体路线见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n## 目标\n\n日志守卫是 `lbs test report` 的一部分，用来在 agent 执行测试期间捕获 UI 断言之外的运行时问题。\n\n当前命令方向已收敛为 `lbs test plan` / `lbs test report`。日志守卫服务于 agent-browser QA，不是独立的后端 API 测试入口。\n\nLangBot 是异步且集成度高的系统，有些问题不会直接表现为页面失败：\n\n- 后台任务异常\n- 未等待的协程\n- Provider 流式调用失败\n- 插件 runtime 超时\n- 平台发送失败\n- 数据库连接问题\n- 敏感信息泄露\n\n日志守卫负责把这些信号结构化地放进测试报告，并关联到 troubleshooting 资产。\n\n## 输入\n\n日志守卫应从环境和运行上下文读取配置：\n\n- `skills/.env` 中的 `LANGBOT_BACKEND_URL`\n- `skills/.env` 中的 `LANGBOT_REPO`，用于自动发现 LangBot 后端日志\n- `lbs test plan` / report 记录的 case id\n- LangBot 后端进程输出\n- 前端 dev server 输出\n- 浏览器 console/network 错误\n- case 声明的 success/failure patterns 和 expected failures\n\n## MVP 范围\n\n- 读取一个或多个日志流或日志文件。\n- 检测错误模式。\n- 支持按 case id 或 pattern 白名单。\n- 输出 JSON/Markdown 摘要。\n- 发现非预期错误时让测试报告标记失败；未来如果有自动执行器，再返回非零退出码。\n\n## 错误分类\n\n### 永远非预期\n\n除非 case 明确声明，否则应失败：\n\n- `Traceback`\n- `Task exception was never retrieved`\n- `RuntimeWarning: coroutine .* was never awaited`\n- `Unclosed client session`\n- `Unclosed connector`\n- `KeyError`\n- `TypeError`\n- `AttributeError`\n- 密钥、token、secret 明文泄露\n\n### Case 预期错误\n\n只有当前 case 声明时允许：\n\n- 无效 provider key\n- Provider 认证失败\n- 无效 webhook payload\n- 插件测试故意抛错\n- 超时测试\n- 限流测试\n\n### 仅警告\n\n报告但默认不失败：\n\n- 可恢复重试\n- 恢复的超时\n- 废弃配置\n- 慢请求\n- 版本检查失败\n\n## 与 Troubleshooting 集成\n\n日志守卫不只输出错误文本，还应尽量匹配已知 troubleshooting id。\n\n例子：\n\n```text\nAction list_plugins call timed out\nAction list_runners call timed out\nAction invoke_llm_stream call timed out\n```\n\n可映射到：\n\n```text\nplugin-runtime-timeout\n```\n\n```text\nuppercase proxy points to one host, lowercase proxy points to another\n```\n\n可映射到：\n\n```text\nproxy-env-mismatch\n```\n\n## 未来命令\n\n```bash\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report --output report.md\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log --console-log /path/to/console.log\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs test report pipeline-debug-chat --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\" --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --no-auto-log\n```\n\n运行报告应包含：\n\n- case id\n- URL 和环境变量摘要，不能包含 secrets\n- 浏览器可见结果\n- 后端日志摘要\n- console/network 错误\n- 匹配到的 troubleshooting id\n- 通过/失败结论\n\n## MVP 完成标准\n\n- 可以自动扫描最新 LangBot 后端日志，也可以扫描前端日志和 console 日志文件。\n- 可以用 `--since` 或 `--tail-lines` 把扫描范围限制到本次测试窗口。\n- 可以检测明显 Python/运行时错误和 secret 泄露风险。\n- 可以识别 case 声明的 success/failure patterns。\n- 可以识别 troubleshooting pattern，包括 `plugin-runtime-timeout` 和 `proxy-env-mismatch`。\n- 支持 case 级白名单。\n- 输出机器可读摘要。\n- 至少一个 `langbot-testing` case 使用它。\n\n当前 MVP 已覆盖自动发现 LangBot 后端日志、文件扫描、`--since`/`--tail-lines` 扫描窗口、\n基础错误检测、case success/failure signal、troubleshooting 匹配、secret 脱敏和 `--json`\n输出。仍待继续完善的是 live log 采集、更多规则、case 级 expected failure 的资产化和真实\nE2E report 样例。\n","skills/qa-agent-docs/qa-agent/03-agent-browser-qa-principles.md":"# Agent Browser QA Principles\n\nThis document fixes the direction of LangBot agent testing so the project does not drift into a backend API smoke-test framework.\n\n## Primary Goal\n\n`langbot-skills` should help an agent behave like a QA engineer using the product, not like a backend curl script.\n\nThe primary path is:\n\n```text\ndeveloper intent -> lbs test plan -> agent controls browser -> UI result + console + logs -> report/assets\n```\n\n## Rules\n\n1. Browser/UI interaction is the source of truth for product QA cases.\n2. A backend API or curl response is never enough to mark a UI case passed.\n3. API/curl/log checks are allowed as diagnostics after a UI path is attempted or when debugging environment readiness.\n4. A case passes only when the user-visible UI result is correct.\n5. The agent should inspect browser console/network output when available.\n6. If screenshot or vision capability is available, the agent should check for blank pages, overlap, hidden actions, broken layout, and error toasts.\n7. If no visual model is available, use DOM/accessibility snapshots and console output instead.\n8. New stable UI paths should be added as `cases/*.yaml`.\n9. New recurring failure modes should be added as `troubleshooting/*.yaml`.\n10. Secrets, tokens, API keys, and localStorage token values must never be printed.\n\n## Command Semantics\n\n`lbs` manages assets and produces plans. It does not replace the agent's browser-control ability.\n\n```bash\nbin/lbs test plan pipeline-debug-chat\n```\n\nThis command outputs:\n\n- environment variables to use\n- required skills\n- browser steps\n- UI/console/visual/log checks\n- diagnostic options\n- related troubleshooting patterns\n- report template\n\nThe active agent then executes the plan with Computer Use, Playwright MCP, or another available browser-control tool.\n\n## Diagnostics\n\nDiagnostics can include:\n\n- `bin/lbs env doctor`\n- browser console/network inspection\n- backend logs\n- targeted API/curl checks\n\nDiagnostics answer \"where did it fail?\" They do not replace \"did the user-visible UI work?\"\n","skills/qa-agent-docs/qa-agent/04-black-box-e2e-roadmap.md":"# 黑盒 E2E QA 路线图\n\n## 定位\n\nLangBot 有大量外部依赖：模型供应商、plugin runtime、浏览器登录态、\nmarketplace/network、RAG engine、sandbox backend、平台适配器等。单测仍然有价值，\n但这个 QA 方向当前不优先解决 LangBot core 的单测覆盖率问题，因为重 mock 往往不能\n真实代表产品路径。\n\n`langbot-skills` 当前目标是让黑盒 E2E 测试变得可执行、可沉淀、可复用：\n\n```text\n开发者测试意图\n-> 复用或新增 case\n-> agent 通过浏览器执行\n-> UI + console + network + log 证据\n-> report\n-> 反哺 case / troubleshooting\n```\n\n这是面向开发者的 QA 资产库。开发者可以让 agent 测一个 feature；如果路径稳定，\n就把路径正规化为 case，让下一个开发者或 QA agent 继续复用。\n\n## 非目标\n\n- 这一阶段不优先建设 LangBot core 单测覆盖率。\n- 不把 API/curl 作为 WebUI 行为的通过标准。\n- 不要求每个 case 都能进 CI。\n- 不在 report 和日志守卫有用之前急着做完整 browser runner。\n- 不把外部 provider、OAuth、marketplace 抖动直接判成产品失败，除非证据明确。\n\n## 当前状态\n\n仓库已经具备第一层基础设施：\n\n- `skills/.env` 和 `skills/.env.local` 管理测试环境；\n- `langbot-env-setup`、`langbot-testing`、`langbot-plugin-dev` 等 skill；\n- `skills/langbot-testing/cases` 下的结构化 case；\n- `skills/langbot-testing/troubleshooting` 下的结构化故障资产；\n- RAG、多模态、plugin、MCP 等 fixture；\n- `bin/lbs validate`、`bin/lbs index`、`bin/lbs case`、`bin/lbs trouble`、\n `bin/lbs test plan`、`bin/lbs test start`、`bin/lbs test report`。\n\n所以当前已经不是“先把路径写进 Markdown”的阶段，而是进入“让每次运行有证据、\n有报告、能沉淀”的阶段。\n\n## 测试模型\n\nUI case 只有在用户可见行为正确时才能通过。辅助证据必须解释同一次运行。\n\n通过一个 UI case 的最低证据：\n\n- 用户可见的成功信号，例如 bot 回复、provider 保存成功、文件上传完成、plugin 页面渲染；\n- 没有意外 browser console error；\n- 相关时间窗口内没有意外后端/runtime 错误；\n- 有截图、DOM snapshot 或同等视觉/结构证据，如果当前 agent 能获取；\n- API/curl 只在解释同一条 UI 路径时作为诊断证据。\n\n失败报告需要保留足够信息，让开发者能复现或分流：\n\n- case id 和实际测试 URL；\n- 使用的 browser path；\n- 最后可见 UI 状态；\n- console/network 症状；\n- 相关后端/前端日志；\n- 匹配到的 troubleshooting id；\n- 这是产品失败、环境问题、外部依赖抖动，还是证据不足。\n\n## 结果词汇\n\n统一使用这些结果：\n\n- `pass`：UI 行为正确，辅助证据干净。\n- `fail`：UI 行为错误，或同一次运行的 console/log 出现意外产品错误。\n- `blocked`：缺登录、缺 provider credentials、服务未启动等原因导致目标路径没有跑起来。\n- `env_issue`：失败在目标行为之外，例如 proxy、OAuth、provider quota、marketplace outage、\n 本地服务启动问题。\n- `flaky`：同一环境下结果不稳定，进入门禁前需要先稳定。\n\n做 merge/release 判断时，`env_issue` 和 `blocked` 不能算产品通过。\n\n## 路线图\n\n### Phase 0：对齐文档\n\n目标：明确当前黑盒 E2E 方向。\n\n交付物：\n\n- `docs/qa-agent/README.md` 文档状态导航；\n- 本路线图；\n- 给旧规划文档加状态说明。\n\n完成标准：\n\n- 新贡献者不用通读所有旧文档，也能知道当前重点。\n\n### Phase 1：Test Report MVP\n\n状态：已有第一版。\n\n目标：让每次 agent browser 测试都有一致报告格式，即使 browser 执行还没自动化。\n\n建议命令：\n\n```bash\nbin/lbs test start \nbin/lbs test report --output reports/-.md\n```\n\nMVP 行为：\n\n- 读取 case 和关联 troubleshooting；\n- 生成 Markdown report 模板；\n- 生成 run handoff，固定本次测试的 start timestamp 和推荐 report command；\n- 写入脱敏后的环境摘要；\n- 提供 `pass/fail/blocked/env_issue/flaky` 结果选项；\n- 包含 UI result、console errors、network symptoms、logs、screenshots、\n diagnostics、matched troubleshooting、assets to update 等 section；\n- 支持 `--json`，输出机器可读报告。\n\n第一版已经是 report generator，不急着做自动判定。先把 evidence 收集格式统一起来，\n再做自动化更稳。\n\n完成标准：\n\n- agent 可以先跑 `lbs test start `，用它给出的时间窗口执行浏览器路径，\n 然后按固定格式填写 report，不需要每次重新发明报告结构。\n\n### Phase 2：日志守卫 MVP\n\n状态：已有第一版文件扫描。\n\n目标：捕获 UI 不一定明显展示的 runtime 问题。\n\n日志守卫应集成进 `lbs test report`，不要发展成独立后端 API 测试框架。\n\n建议命令形态：\n\n```bash\nbin/lbs test report \\\n --backend-log /path/to/backend.log \\\n --frontend-log /path/to/frontend.log \\\n --console-log /path/to/console.log \\\n --evidence-dir reports/evidence/ \\\n --since \"2026-05-21T10:30:00+08:00\" \\\n --tail-lines 2000 \\\n --output reports/-.md\n```\n\nMVP 行为：\n\n- 默认从 `LANGBOT_REPO/data/logs/` 扫描最新 `langbot-*.log`；\n- 支持 agent 显式提供 backend、frontend、console 日志文件；\n- 支持读取 evidence 目录下的 `automation-result.json`，把浏览器自动化脚本结论纳入报告；\n- 支持 `lbs test result` 为人工/agent browser 运行写入标准 `result.json`，供 suite 聚合；\n- 支持 `--since` 和 `--tail-lines`，避免历史日志污染本次报告；\n- 检测默认非预期模式，例如 `Traceback`、未 await coroutine、unclosed client/connector、\n `KeyError`、`TypeError`、`AttributeError`、明显 secret 泄露；\n- 匹配 case 声明的 `success_patterns` 和 `failure_patterns`；\n- 匹配已知 troubleshooting，先支持 `plugin-runtime-timeout` 和 `proxy-env-mismatch`；\n- 只有 case 明确声明时，才允许 expected failure；\n- 将发现分类为 fail、warning、matched troubleshooting、ignored expected issue；\n- 永远不打印 secret 值。\n\n完成标准：\n\n- 至少 `pipeline-debug-chat` 能生成包含日志摘要和 troubleshooting 匹配结果的 report。\n\n### Phase 3：Case 元数据加固\n\n状态：已有第一版。\n\n目标：让 case 更容易选择、执行和晋级。\n\n字段逐步补充，保持向后兼容：\n\n```yaml\npriority: p0 | p1 | p2\nrisk: low | medium | high\nci_eligible: false\npreconditions:\n - \"Authenticated browser profile is available.\"\nsetup:\n - \"Start LangBot backend and frontend.\"\ncleanup:\n - \"Remove temporary provider, plugin, or knowledge base if created.\"\nexpected_failures: []\nsuccess_patterns:\n - \"Conversation(0) Streaming completed\"\nfailure_patterns:\n - \"Action invoke_llm_stream call timed out\"\nevidence:\n required:\n - ui\n - console\n - backend_log\n```\n\n当前实现采用扁平字段 `evidence_required`，避免轻量 YAML 解析器在 case 文件里承载嵌套结构。\n`bin/lbs validate` 会校验 `priority`、`risk`、`ci_eligible`、`evidence_required`、\n`automation` 脚本路径、case 关联 skill 和 troubleshooting 交叉引用。`bin/lbs case list`\n支持 `--json`、`--type`、`--area`、`--tag`、`--priority`、`--risk`、`--automation`、`--ci`\n、`--ready` 和 `--machine-ready` 过滤，方便 agent 快速选择测试集。\n`env_any` 和 `automation_env_any` 用于表达 URL-or-name 这类 one-of 输入，避免把可替代变量误判为全部必填。\n\n当前也有 `skills//suites/*.yaml` 和 `bin/lbs suite plan `，用于组织常跑测试集，\n例如 `core-smoke`、`local-agent-gate` 和\n`agent-runner-release-gate`。发布门禁使用 `agent-runner-release-preflight`\n先分类配置 blockers 和 runtime env issues，再运行较重的浏览器 Debug Chat case。\n依赖 fixture 的 case 可以在浏览器执行前先跑 `bin/lbs fixture check`，检查\n`fixtures/fixtures.json` 登记的 deterministic 文件、plugin 包和本地测试 server 是否存在。\n`bin/lbs suite start ` 会生成 suite run id、suite evidence root、per-case evidence 目录、\n`suite-start.json`/`suite-start.md` handoff 文件和 per-case evidence 命令；\n浏览器自动化脚本会写入 `automation-result.json`，供 `bin/lbs test report` 展示原始自动化结论；\n`bin/lbs test result ` 会在人工/agent browser case 完成后写入最终 `result.json`；\n`bin/lbs suite report --evidence-dir ` 会聚合各 case 的 `result.json`，并且\n不会把缺少 required evidence 的 `pass` 当作 suite 通过。\nRunner 专用 Debug Chat case 通过 `automation_pipeline_url_env` 和\n`automation_pipeline_name_env` 绑定专用 pipeline 变量，避免 local-agent、Codex 或\nClaude Code case 误用通用 `LANGBOT_PIPELINE_URL` 后产生假阳性。\nDebug Chat case 还可以通过 `automation_stream_output` 固定流式或非流式发送路径。\n多模态 Debug Chat case 可以通过 `automation_image_base64_fixture` 复用 deterministic 图片 fixture。\n`test plan` 和 `suite plan` 会输出 readiness，让 agent 在执行浏览器前就看到缺失的 env、\n自动化变量、fixture，以及需要人工确认的 `manual_check` 前置条件。\n\n完成标准：\n\n- `lbs case list` 或后续 filter 能回答“smoke 跑哪些”、“哪些适合 CI”、\n “哪些需要真实 provider credentials”。\n\n### Phase 4：开发者沉淀流程\n\n目标：开发者让 agent 测新 feature 后，稳定路径不会丢在聊天记录里。\n\n流程：\n\n1. 开发者要求 agent 通过浏览器测试某个 feature。\n2. agent 先按 UI 主路径探索。\n3. agent 用 `lbs test start` 固定运行窗口，再用 `lbs test report` 写报告。\n4. 如果路径稳定，agent 新增或更新 case。\n5. 如果出现可复用故障，agent 新增或更新 troubleshooting。\n6. agent 跑 `bin/lbs validate` 和 `bin/lbs index`。\n\n完成标准：\n\n- feature QA 的结果能进入资产库，而不是只留在一次对话里。\n\n### Phase 5：选择性浏览器自动化\n\n状态：已有第一版 `test run` 入口和两个 Playwright 脚本。\n\n目标：只自动化少量稳定、值得重复跑的黑盒路径。\n\n建议顺序：\n\n1. `webui-login-state`\n2. `pipeline-debug-chat`\n3. `local-agent-basic-debug-chat`\n4. `local-agent-rag-debug-chat`\n5. 一个基于 deterministic fixture 的 plugin 或 MCP smoke path\n\n执行策略：\n\n- 继续把 Computer Use 或 Playwright MCP 作为默认交互路径；\n- 只给稳定、确定性的路径补直接 Playwright script；\n- 保存 screenshots、console logs、trace/video；\n- flaky 或强依赖真实 credentials 的 provider case 暂时不要进 CI。\n\n当前已经绑定：\n\n- `webui-login-state` -> `scripts/e2e/webui-login-state.mjs`\n- `pipeline-debug-chat` -> `scripts/e2e/pipeline-debug-chat.mjs`\n\n第一版自动化先产出 `reports/evidence//` 下的 console、network、screenshot 和\nresult JSON。真实执行后仍要用 `lbs test report --since ... --console-log ...` 做日志守卫和\n最终报告。开发期间可以先用 `bin/lbs test run --dry-run` 检查命令和 evidence 路径。\nDebug Chat 类脚本应复用 `scripts/e2e/lib/debug-chat.mjs`，避免重复实现 visible response leaf\n判断和已知失败信号分类。\n\n完成标准：\n\n- 小规模 smoke subset 可以不靠人工决定每一步点击；更大的资产库仍然服务于人工/agent\n 驱动的探索式 E2E。\n\n## 下一批动工切片\n\n在做 browser runner 之前，继续做这些：\n\n1. 等 LangBot 当前开发状态稳定后，用一次真实 `pipeline-debug-chat` 跑通\n `test start -> test run -> test report -> test result -> suite report`，产出 sample report。\n2. 只给 smoke/local-agent 首批 case 补必要元数据。\n3. 继续补日志守卫规则，尤其是 WebSocket、plugin runtime、provider streaming、前端\n chunk/rendering failure。\n4. 约定 report 产物目录、截图和 console/network 导出的命名方式。\n5. 再评估是否开始给 `webui-login-state` 和 `pipeline-debug-chat` 做直接 Playwright\n 自动化。\n\n这样 infra 会立刻有用，同时保留后续自动化 browser execution 的空间。\n","skills/qa-agent-docs/qa-agent/README.md":"# LangBot QA Agent 文档导航\n\n这个目录记录 `langbot-skills` 当前的 QA 方向和后续建设顺序。\n\n## 当前判断\n\n当前重点是 LangBot 的黑盒 E2E QA，不是 LangBot core 的单测覆盖率建设。\n\n`langbot-skills` 要帮助开发者和 QA agent 做接近人工测试的 WebUI 验证：\n\n- 打开真实 LangBot WebUI；\n- 按用户路径点击和输入；\n- 检查用户可见的 UI 结果；\n- 查看 console、network、截图、后端和前端日志；\n- 输出可复用的测试报告；\n- 把稳定 feature 路径沉淀为 case；\n- 把重复故障沉淀为 troubleshooting。\n\nAPI 和 curl 只做诊断。它们可以解释失败原因，但不能让一个 UI case 通过。\n\n## 文档状态\n\n| 文档 | 状态 | 用途 |\n| --- | --- | --- |\n| `04-black-box-e2e-roadmap.md` | 当前主路线图 | 决定下一步建设什么。 |\n| `03-agent-browser-qa-principles.md` | 当前原则文档 | 定义 browser-first QA 的通过标准。 |\n| `02-log-guard-plan.md` | 当前活跃设计 | 设计 `lbs test report` 里的日志守卫。 |\n| `../user-guide.md` | 当前使用手册 | 开发者日常使用。 |\n| `00-technology-options.md` | 背景文档 | 选择 Computer Use、Playwright MCP 或未来直接 Playwright。 |\n| `01-qa-agent-harness-plan.md` | 历史规划，部分过时 | 解释最初分层和目录设计；使用前先看状态说明。 |\n\n## 已过时的点\n\n`01-qa-agent-harness-plan.md` 还保留早期规划状态。现在结构化 cases、\n结构化 troubleshooting、`validate`、`index`、`lbs test plan` 都已经落地。\n\n已经补上第一版 `lbs test start`、`lbs test run`、`lbs test report` 和日志守卫文件扫描。\n`webui-login-state`、`pipeline-debug-chat` 已经绑定直接 Playwright 自动化脚本。后续重点是：\n\n- 报告 evidence 字段继续打磨；\n- case success/failure signal 和日志守卫规则继续补充；\n- 报告产物和 evidence 约定；\n- 等 LangBot 当前开发状态稳定后跑真实 sample report。\n\n不要再把旧阶段列表当成当前 source of truth。后续排序以\n`04-black-box-e2e-roadmap.md` 为准。\n","skills/qa-agent-docs/user-guide.md":"# LangBot Skills 用户使用手册\n\n## 这个仓库解决什么\n\n`langbot-skills` 是给 agent 使用的 LangBot 测试资产库。开发者 clone 后，可以让 Codex、Claude Code、Computer Use 或 Playwright MCP 复用已有环境配置、测试路径和故障知识，像 QA 一样操作 LangBot WebUI。\n\n核心目标：\n\n- 不让下一个 agent 窗口从头探索登录、模型配置、pipeline 调试。\n- 把稳定 UI 测试路径沉淀为 case。\n- 把常见故障沉淀为 troubleshooting。\n- 让 agent 优先通过浏览器点击验证产品行为。\n- API/curl/log 只作为诊断手段，不作为 UI case 通过标准。\n\n## 快速开始\n\n1. Clone 仓库。\n\n2. 检查本地默认变量：\n\n```bash\nbin/lbs env show\n```\n\n默认变量在：\n\n```text\nskills/.env\n```\n\n本机专用覆盖写到：\n\n```text\nskills/.env.local\n```\n\n它会覆盖 `skills/.env` 中的同名变量，并且不应该提交。\n`skills/.env` 是共享默认值，不应写入本机绝对路径、浏览器 profile、provider key 或其他凭据。\n新机器建议从模板开始：\n\n```bash\ncp skills/.env.example skills/.env.local\n```\n\n常用变量：\n\n```text\nLANGBOT_FRONTEND_URL\nLANGBOT_BACKEND_URL\nLANGBOT_DEV_FRONTEND_URL\nLANGBOT_REPO\nLANGBOT_WEB_REPO\nLANGBOT_BROWSER_PROFILE\n```\n\n3. 检查环境是否就绪：\n\n```bash\nbin/lbs env doctor\nbin/lbs fixture check\n```\n\n`env doctor` 会检查 URL、路径、代理变量等。代理变量是可选项；只有大小写代理变量互相冲突时才会报错。失败不一定代表仓库坏了，通常说明本地 LangBot 没启动、代理不一致或浏览器 profile 不存在。\n`fixture check` 会检查仓库内测试 fixture 是否存在，例如 MCP stdio server、RAG 文档、多模态图片、qa-plugin-smoke 包和 QA Runner 包。它也会校验 `.lbpkg` 是 zip 包，并检查 QA Runner fixture 的入口文件未漂移。\n\n4. 查看已有测试 case：\n\n```bash\nbin/lbs case list\nbin/lbs case list --json --priority p0 --automation\nbin/lbs case list --ready\nbin/lbs case list --machine-ready\nbin/lbs suite list\nbin/lbs suite plan core-smoke\nbin/lbs suite plan agent-runner-release-gate\nbin/lbs suite start core-smoke\nbin/lbs suite start core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\n```\n\n`case list` 支持按 `--type`、`--area`、`--tag`、`--priority`、`--risk`、`--automation`\n、`--ci`、`--ready` 和 `--machine-ready` 过滤。`--ready` 只显示没有缺机器输入且没有人工前置条件的 case；\n`--machine-ready` 过滤掉缺机器输入的 case，但保留 `manual-check`，表示执行前还要确认前置条件。需要交给 agent 自动选择测试集时，优先使用 `--json`，\n其中包含 `priority`、`risk`、`ci_eligible`、`automation`、`evidence_required` 以及\nenv/automation/fixture/manual readiness。\nCase metadata 中的 `env` 和 `automation_env` 表示全部必填；URL 或 name 这类二选一输入会放在\n`env_any` 或 `automation_env_any`，readiness 只要求组合里至少一个变量有值。\n\n如果要跑一组已沉淀的测试路径，优先使用 suite。Suite 位于 `skills//suites/*.yaml`，\n只负责组织 case，不改变 UI/browser 作为通过标准的原则。\n`suite plan` 会聚合 readiness：缺环境变量、缺自动化变量、缺 fixture 或需要\n`manual_check` 时，会在执行前标出受影响的 case。`manual_check` 不是产品通过，\n它表示机器配置已满足但 agent 必须先确认 case 里的 `preconditions` 或 `setup`。\nRunner externalization 发布判断使用 `agent-runner-release-gate`。先跑\n`agent-runner-release-preflight`，把缺 pipeline、runner id 错误、插件未安装这类\n`blocked`，以及 provider key、Box、插件运行时这类 `env_issue` 分开，再执行较重的\n浏览器 Debug Chat case。\n\n5. 生成 agent 执行计划：\n\n```bash\nbin/lbs test plan pipeline-debug-chat\n```\n\n然后把计划交给当前 agent 执行。agent 应使用 Computer Use、Playwright MCP 或其他浏览器控制能力去操作 UI。\n`test plan` 中的 Environment、Automation Readiness、Fixture Readiness 和 Manual\nReadiness 是执行前门禁；如果 readiness 缺失，应先补配置或将本次 case 标记为\n`blocked`。如果状态是 `manual_check`，先确认 `preconditions` 和 `setup`，再开始 UI\n执行。不要把后续 curl/API 诊断当成 UI case 通过。\n\n## 推荐使用方式\n\n### 冒烟测试\n\n你可以直接对 agent 说：\n\n```text\n帮我跑一下 LangBot 新前端 smoke test。\n```\n\nagent 应该：\n\n- 读 `skills/.env`\n- 优先查看 `bin/lbs suite plan core-smoke`，或查找 `type: smoke` 的 cases\n- 生成 test plan\n- 用浏览器执行 UI 操作\n- 检查 console、截图、后端日志\n- 输出简短 QA 报告\n\n### Runner Externalization 发布门禁\n\n你可以直接对 agent 说：\n\n```text\n按 agent-runner release gate 跑完整矩阵，先做 preflight，再跑浏览器 case，并把 blocked/env_issue/fail 分开。\n```\n\nagent 应该先查看 `skills/langbot-testing/references/agent-runner-release-gate.md`，\n再执行：\n\n```bash\nbin/lbs test recommend\nbin/lbs suite plan agent-runner-release-gate\nbin/lbs test run agent-runner-release-preflight --dry-run\nbin/lbs suite start agent-runner-release-gate --run-id agent-runner-release-local --evidence-dir reports/evidence/agent-runner-release-local\n```\n\n`test recommend` 输出的 run 命令默认带 `--dry-run`；确认 readiness 和 `manual_check` 前置条件后，再去掉 `--dry-run` 执行。\n\n完成所有 case 后，用：\n\n```bash\nbin/lbs suite report agent-runner-release-gate --evidence-dir reports/evidence/agent-runner-release-local\n```\n\n没有最终 `result.json`、缺 required evidence、或把 `blocked`/`env_issue` 当成 pass，\n都不能算发布门禁通过。\n\n### 新 Feature 测试\n\n你可以说：\n\n```text\n我改了 provider 页面，帮我测一下 DeepSeek provider 添加、测试、绑定 pipeline 是否正常。\n```\n\nagent 应该：\n\n- 查找相关 case 和 reference\n- 如果没有稳定路径，先探索 UI\n- 用浏览器执行真实交互\n- 失败时用日志/API 辅助定位\n- 稳定后新增或更新 case/reference\n- 新故障沉淀为 troubleshooting\n\n### 定点排错\n\n你可以说：\n\n```text\nDebug Chat 点了没回复，帮我查是前端问题还是后端问题。\n```\n\nagent 应该：\n\n- 先通过 UI 复现\n- 看 console/network\n- 看后端日志\n- 必要时用 API/curl 做诊断\n- 匹配 troubleshooting\n- 给出修复建议或直接修复\n\n## 重要原则\n\n这些原则固定在：\n\n```text\ndocs/qa-agent/03-agent-browser-qa-principles.md\n```\n\n简化版：\n\n- UI/browser 是测试主路径。\n- API/curl/log 只做诊断。\n- 后端接口成功不等于 UI case 通过。\n- case 通过必须以用户可见 UI 结果为准。\n- 有视觉能力时应检查截图。\n- 没有视觉能力时用 DOM/accessibility snapshot 和 console。\n- 不要打印 token、API key、OAuth secret 或 localStorage token 值。\n\n## 规划文档\n\n如果要判断下一步建设什么，先看：\n\n```text\ndocs/qa-agent/README.md\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n`01-qa-agent-harness-plan.md` 是早期规划，部分内容已经被当前实现和路线图替代。\n\n## 常用命令\n\n### 环境\n\n```bash\nbin/lbs env show\nbin/lbs env show --json\nbin/lbs env doctor\nbin/lbs fixture list\nbin/lbs fixture check\nbin/lbs fixture check --json\n```\n\n`env show` 和 `env doctor` 默认会对 token、API key、password、secret 以及 URL basic auth\n做脱敏。不要把 `.env.local` 里的原始凭据复制进测试报告。\n\n### Skill 和索引\n\n```bash\nbin/lbs list\nbin/lbs validate\nbin/lbs index --check\nbin/lbs index\n```\n\n### Case\n\n```bash\nbin/lbs case list\nbin/lbs case list --type smoke\nbin/lbs case list --json --priority p1 --tag local-agent\nbin/lbs case list --ready\nbin/lbs case list --machine-ready\nbin/lbs case show pipeline-debug-chat\nbin/lbs case new my-feature --title \"My Feature Works\"\n```\n\n### Suite\n\n```bash\nbin/lbs suite list\nbin/lbs suite list --json --priority p1\nbin/lbs suite show local-agent-gate\nbin/lbs suite plan core-smoke\nbin/lbs suite plan local-agent-gate --json\nbin/lbs suite start core-smoke\nbin/lbs suite start core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\nbin/lbs suite run core-smoke --dry-run --json\nbin/lbs suite run core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\nbin/lbs suite start core-smoke --json\nbin/lbs suite report core-smoke --evidence-dir reports/evidence/\nbin/lbs suite report core-smoke --evidence-dir reports/evidence/ --json\nbin/lbs suite new my-feature-gate --title \"My Feature Gate\"\n```\n\n`suite start` 不直接控制浏览器。它生成统一 run id、suite evidence root、每个 case 的 evidence\n目录、`suite-start.json`/`suite-start.md` handoff 文件，以及每个 case 的 `test run`、`test report`\n和 `test result` 命令模板。需要固定路径时，使用 `--run-id` 和 `--evidence-dir`。\n`suite run --dry-run --json` 只预览 planned/skipped case，不创建 evidence，也不执行 automation。\n`suite run` 会顺序执行 suite 中已有 automation、机器 readiness 已满足且不需要 `manual_check` 的 case，并在最后聚合 `suite report`。\n缺 env、automation env 或 fixture 的 case 默认会跳过；确实要强制执行时，加 `--include-not-ready`。\n确认前置条件后，才用 `--include-manual-check` 执行这类 case。\n所有 case 执行完并写入最终 `result.json` 后，\n`suite report` 会读取各 case evidence 目录并汇总为 `pass`、`fail`、`blocked`、`env_issue`、\n`flaky`、`incomplete` 等状态。`pass` 必须声明已经收集 case 的全部 required evidence；\n否则 suite 会保持 `incomplete`，避免把缺证据的运行误判成通过。\n\n### Test Plan\n\n```bash\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test plan pipeline-debug-chat --json\n```\n\n### Test Start\n\n```bash\nbin/lbs test start pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat --json\n```\n\n`test start` 用于 agent 开始一次浏览器测试前记录 run id、开始时间和推荐 report 命令。\n它会把 `--since \"\"` 写进后续报告命令，减少历史日志污染本次判断。\n如果 case 绑定了自动化脚本，输出里也会包含 `test run` 命令和 evidence 目录。\n\n### Test Automation\n\n```bash\nbin/lbs test run webui-login-state --dry-run\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test run webui-login-state --run-id login-smoke --output reports/evidence/login-smoke\nbin/lbs test run pipeline-debug-chat --run-id pipeline-smoke --output reports/evidence/pipeline-smoke\n```\n\n查看当前所有带自动化脚本的 case：\n\n```bash\nbin/lbs case list --automation\nbin/lbs case list --json --automation\n```\n\n当前自动化覆盖包括登录态、通用 Pipeline Debug Chat、local-agent runner 的基础回复、\nPromptPreProcessing、RAG、plugin tool、MCP stdio tool、非流式、多模态和 RAG+多模态路径。\n不要在文档里手工维护静态 case 清单；以 `case list --automation` 和 suite 定义为准。\n\n自动化脚本位于 `scripts/e2e/`。它们会保存：\n\n- `console.log`\n- `network.log`\n- `screenshot.png`\n- `automation-result.json`\n\n新增 Debug Chat 类自动化时，优先复用 `scripts/e2e/lib/debug-chat.mjs` 中的 pipeline 打开、\nprompt 发送、visible response leaf 判断和失败信号分类，不要在新脚本里复制 DOM 扫描逻辑。\n\n脚本需要本地安装 Playwright 后才能真正执行：\n\n```bash\nnpm install\nnpx playwright install chromium\n```\n\n`pipeline-debug-chat` 通用自动化建议配置 `LANGBOT_PIPELINE_URL`。如果没有 direct URL，\n脚本会尝试通过 `LANGBOT_PIPELINE_NAME` 从 Pipelines 页面进入目标 pipeline。两者都没有时，\n该自动化会返回 `blocked`，不会伪造通过。\n\nRunner 专用 case 不应复用通用 pipeline 变量。Local Agent、Codex Runner 和\nClaude Code Runner 这类 case 会通过 `automation_pipeline_url_env` /\n`automation_pipeline_name_env` 映射到 case-specific env，例如\n`LANGBOT_LOCAL_AGENT_PIPELINE_URL`。这些 case 如果缺少专用变量，会返回 `blocked`，\n不会退回到 `LANGBOT_PIPELINE_URL`，避免跑错 pipeline 后产生假阳性。\n如果 case 声明了 `setup_automation`，只有 `bin/lbs test run ` 的真实执行路径会先运行这些 setup；\n`test plan`、`suite plan`、`case list` 和 `--dry-run` 只展示它们，不会修改本地环境。\nsetup 可以是 `case:` 或仓库内 `node:scripts/... --flag`，每一步证据会写到主 evidence 目录下的\n`setup/` 子目录。setup 失败时主 automation 不会继续执行；setup 写入 `.env.local` 后，主 automation\n会重新读取环境。用 `setup_provides_env` 声明 setup 会生成的变量，可以让 readiness 正确显示机器可准备状态。\n如果 Debug Chat case 需要固定流式/非流式路径，可以在 case 中设置\n`automation_stream_output: \"1\"` 或 `\"0\"`，脚本会在发送消息前切换 Debug Chat 的 Stream 控件。\n如果 case 需要上传图片，可以设置 `automation_image_base64_fixture` 指向仓库内的 base64 PNG fixture，\n脚本会在 evidence 目录写出临时 PNG 并通过 Debug Chat 上传控件发送。\n`bin/lbs test plan --json` 和 `bin/lbs suite plan --json`\n都会显示这些专用变量是否已配置。\n\n### Test Report 和日志守卫\n\n```bash\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report pipeline-debug-chat --output reports/pipeline-debug-chat.md\nbin/lbs test report pipeline-debug-chat \\\n --backend-log /path/to/backend.log \\\n --frontend-log /path/to/frontend.log \\\n --console-log /path/to/console.log\nbin/lbs test report pipeline-debug-chat --evidence-dir reports/evidence/pipeline-smoke\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log --json\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs test report pipeline-debug-chat --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\" --tail-lines 2000\n```\n\n`test report` 会生成报告模板，并默认从 `LANGBOT_REPO/data/logs/` 自动选择最新的\n`langbot-*.log` 作为 LangBot 后端日志扫描。也可以用 `--backend-log` 覆盖，或用\n`--no-auto-log` 只生成模板。\n\n如果提供 `--evidence-dir`，或 `--console-log` 指向 `reports/evidence//console.log`，\n报告会优先读取同目录的 `automation-result.json`，并展示自动化脚本的 `status`、`reason`、\n起止时间和目标 URL。\n\n日志守卫会扫描常见错误、secret 泄露风险、case 声明的 success/failure patterns，以及已知\ntroubleshooting pattern。它不控制浏览器，也不替代 UI 通过判断。`success_patterns`\n命中会作为通过证据写入 `success_signals`；声明了 success pattern 但本次扫描窗口没有命中，\n会给 warning；`failure_patterns` 命中会让本次日志守卫 fail。\n\n建议在执行浏览器 case 前记录当前时间，然后在报告阶段使用 `--since`。如果只想快速看\n最近日志，可以使用 `--tail-lines`。\n\n### Runtime Log Guard\n\n如果还没有进入某个具体 UI case，只是想观察 LangBot 后端日志，可以直接使用 `log`\n命令。它和 `test report` 使用同一套扫描器、secret 脱敏、troubleshooting pattern 和\ncase success/failure pattern。\n\n```bash\nbin/lbs log scan --tail-lines 300\nbin/lbs log scan --case pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs log scan --backend-log /path/to/langbot.log --json\nbin/lbs log scan --failure-pattern \"runner.tool_loop_error|Action invoke_llm_stream call timed out\" --strict\n```\n\n`log scan` 默认从 `LANGBOT_REPO/data/logs/` 自动选择最新的 `langbot-*.log`。传入\n`--case ` 后，会额外应用该 case 声明的 `success_patterns`、`failure_patterns`\n和 related troubleshooting。默认用于观察，返回码保持 0；加 `--strict` 后，`fail` 或\n`env_issue` 会返回非 0，适合脚本门禁。\n\n运行期观察可以用 `watch`：\n\n```bash\nbin/lbs log watch --case pipeline-debug-chat\nbin/lbs log watch --backend-log /path/to/langbot.log --interval-ms 500\nbin/lbs log watch --duration-ms 30000 --strict --json\n```\n\n`log watch` 默认从启动时的文件末尾开始，只观察新追加的日志；加 `--from-start` 可从文件开头扫。\n它会实时打印新命中的 findings 和 success signals。为了避免当前历史日志噪声影响观察，默认不因\n异常返回非 0；加 `--strict` 后，退出时如果看到 `fail` 或 `env_issue` 会返回非 0。\n\n给一次 QA 运行包日志窗口时，用 `guard start/stop`：\n\n```bash\nbin/lbs log guard start --run-id local-debug --case pipeline-debug-chat\n# 执行浏览器或手工测试\nbin/lbs log guard stop --run-id local-debug\n```\n\n`start` 会在 `reports/log-guards/.json` 记录起始时间、case 和当前后端日志路径；\n`stop` 会用 start/stop 时间作为扫描窗口，生成 `reports/log-guards/.md`，并默认按\nstrict guard 返回码处理。临时只想收集报告、不想让命令失败，可以加 `--no-strict`。\n\n当前 LangBot core 日志还不是完全结构化日志，runtime guard 主要依赖时间窗口和文本 pattern。\n已支持 ISO 时间戳和 LangBot 当前的 `[MM-DD HH:mm:ss.SSS]` 前缀；没有时间戳的连续行会跟随上一条\n带时间戳的日志块。如果后续 core 能提供稳定 request id、conversation id、plugin action id 或\nJSON log field，guard 可以从“时间窗口 + 文本匹配”升级为更精确的关联分析。\n\n### Test Result\n\n```bash\nbin/lbs test result pipeline-debug-chat \\\n --result pass \\\n --reason \"Debug Chat returned OK and the report log guard was clean.\" \\\n --evidence-dir reports/evidence/pipeline-smoke \\\n --started-at \"2026-05-21T10:30:00+08:00\" \\\n --evidence ui,screenshot,console,backend_log\n```\n\n`test result` 用于把一次人工/agent browser 运行的最终判断写成标准 `result.json`，\n供 `suite report` 聚合。它不会替代 UI 测试：如果写 `--result pass`，`--evidence`\n必须覆盖该 case 的 `evidence_required`，否则命令会失败。自动化脚本写\n`automation-result.json`；如果 case 还要求 backend log、API diagnostic 或 filesystem\nevidence，agent 需要在报告和诊断完成后再用 `test result` 写最终 `result.json`。\n\n### Troubleshooting\n\n```bash\nbin/lbs trouble list langbot-testing\nbin/lbs trouble show plugin-runtime-timeout\nbin/lbs trouble search runtime\nbin/lbs trouble add langbot-testing --title \"...\" --symptom \"...\" --cause \"...\" --fix \"...\"\n```\n\n## 目录说明\n\n```text\nskills/\n .env # 共享默认变量\n langbot-env-setup/ # 环境、浏览器、登录态、代理\n langbot-testing/ # WebUI / provider / pipeline 测试\nschemas/ # 结构化资产 schema\nsrc/ # lbs TypeScript 源码\nbin/ # lbs 入口\ndocs/ # 设计文档和用户手册\nAGENTS.md # agent 维护协议\n```\n\n## 添加一个新测试路径\n\n1. 先让 agent 通过浏览器探索并执行路径。\n2. 稳定后创建 case：\n\n```bash\nbin/lbs case new provider-xxx --title \"Provider XXX can be configured\" --area provider --type provider\n```\n\n3. 编辑生成的 `cases/*.yaml`，补充真实步骤、检查点和 troubleshooting。\n\n4. 校验：\n\n```bash\nbin/lbs validate\nbin/lbs index --check\nbin/lbs index\n```\n\n## 添加一个新故障\n\n```bash\nbin/lbs trouble add langbot-testing \\\n --title \"Plugin runtime actions time out\" \\\n --symptom \"Debug Chat shows Agent runner temporarily unavailable\" \\\n --cause \"Old plugin runtime survived backend restart\" \\\n --fix \"Stop old runtime processes and restart LangBot\"\n```\n\n然后编辑生成的 YAML，补充 `patterns`、`related_cases` 和验证方式。\n\n## 当前边界\n\n- `lbs test plan` 只生成测试计划，不直接控制浏览器。\n- `lbs test report` 生成报告，默认扫描最新 LangBot 后端日志；也可扫描显式提供的\n backend/frontend/console 日志文件。\n- 真正的 UI 操作由当前 agent 的浏览器能力执行。\n- `env doctor` 是 readiness check，不是产品测试。\n- `curl/API` 是诊断工具，不是主要测试路径。\n","skills/schemas/README.md":"# Schemas\n\n这个目录存放 LangBot skills 结构化资产的 JSON Schema。\n\n它们不是测试脚本，也不会执行浏览器动作。它们的作用是定义 agent 和维护者后续新增资产时应该遵守的文件结构。\n\n## 文件说明\n\n- `skills//fixtures/fixtures.json`\n 不是 JSON Schema，但由 `bin/lbs validate` 校验。\n 它登记 deterministic fixture 文件、类型和关联 case，供 `bin/lbs fixture check` 做 readiness 检查。\n\n- `case.schema.json`\n 约束 `skills//cases/*.yaml` 的格式。\n Case 描述 agent-browser 或 probe QA 路径，包括前置条件、步骤、检查点、诊断手段和关联故障。\n\n- `suite.schema.json`\n 约束 `skills//suites/*.yaml` 的格式。\n Suite 只组织 case 集合，用于 smoke、regression 或 release gate 等测试入口。\n\n- `troubleshooting.schema.json`\n 约束 `skills//troubleshooting/*.yaml` 的格式。\n Troubleshooting 条目描述症状、日志/错误模式、可能原因、修复步骤和验证信号。\n\n- `skill-index.schema.json`\n 约束生成文件 `skills.index.json` 的格式。\n 这个索引用于让 agent 快速发现已有 skills、references、cases、suites 和 troubleshooting。\n\n- `reports/evidence//result.json`\n 不是 catalog schema，而是执行期最终裁定产物，由 `bin/lbs test result` 写入。\n `suite report` 读取其中的 `status`、`reason`、起止时间和 `evidence_collected`，\n 并用 `evidence_missing` 防止缺证据的 `pass` 被当作完整通过。\n\n- `reports/evidence//automation-result.json`\n 不是 catalog schema，而是浏览器自动化脚本的原始运行结论，供 `bin/lbs test report`\n 展示和推断日志扫描窗口。\n\n## 为什么需要 schemas\n\nSchemas 是基础设施护栏：\n\n- 防止 case、suite 和 troubleshooting 随着增长变得格式混乱\n- 让 `bin/lbs validate` 能发现缺字段和错误结构\n- 为未来编辑器提示和 CI 校验留接口\n- 帮助 agent 新增资产时知道应该写哪些字段\n\n## 当前校验方式\n\n`bin/lbs validate` 做轻量、schema 对齐的校验，不引入额外依赖。它会检查必填字段、\n枚举值、boolean 字段、重复列表项、automation 脚本存在性，以及 case、suite、skill、\ntroubleshooting 之间的交叉引用。这里的 schema 仍是格式契约；如果未来引入正式 JSON\nSchema validator，应继续保持这些本地交叉引用检查。\n\nCase 里的 `env` / `automation_env` 表示所有列出的变量都需要配置。遇到二选一输入时，\n使用 `env_any` / `automation_env_any`，每一项写成 `LANGBOT_PIPELINE_URL|LANGBOT_PIPELINE_NAME`\n这类 one-of 组合，避免 agent 因为只配置了 URL 或 name 其中之一而误判未就绪。\n`setup` 和 `preconditions` 是人工确认项，会让 readiness 进入 `manual_check`；\n`setup_automation` 是 `test run` 可以自动执行的准备步骤，配合 `setup_provides_env`\n声明它会生成的机器变量。\n","skills/schemas/case.schema.json":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/case.schema.json\",\n \"title\": \"LangBot Skill Test Case\",\n \"type\": \"object\",\n \"required\": [\n \"id\",\n \"title\",\n \"mode\",\n \"area\",\n \"type\",\n \"priority\",\n \"risk\",\n \"ci_eligible\",\n \"tags\",\n \"skills\",\n \"steps\",\n \"checks\",\n \"evidence_required\"\n ],\n \"allOf\": [\n {\n \"if\": {\n \"properties\": {\n \"mode\": { \"const\": \"agent-browser\" }\n }\n },\n \"then\": {\n \"required\": [\"env\"]\n }\n }\n ],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"mode\": {\n \"type\": \"string\",\n \"enum\": [\"agent-browser\", \"probe\"]\n },\n \"area\": {\n \"type\": \"string\"\n },\n \"type\": {\n \"type\": \"string\",\n \"enum\": [\n \"smoke\",\n \"regression\",\n \"feature\",\n \"provider\",\n \"exploratory\",\n \"contract\",\n \"performance\",\n \"reliability\",\n \"chaos\",\n \"security\"\n ]\n },\n \"priority\": {\n \"type\": \"string\",\n \"enum\": [\"p0\", \"p1\", \"p2\"]\n },\n \"risk\": {\n \"type\": \"string\",\n \"enum\": [\"low\", \"medium\", \"high\"]\n },\n \"ci_eligible\": {\n \"type\": \"boolean\"\n },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"skills\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"env_any\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*(\\\\|[A-Z][A-Z0-9_]*)+$\"\n }\n },\n \"steps\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"checks\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"evidence_required\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"enum\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"backend_log\",\n \"frontend_log\",\n \"api_diagnostic\",\n \"filesystem\",\n \"metrics\",\n \"trace\",\n \"profile\",\n \"resource_log\"\n ]\n },\n \"minItems\": 1\n },\n \"preconditions\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup_automation\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^(?:case:[a-z0-9][a-z0-9_-]*|node:scripts/[A-Za-z0-9_./-]+\\\\.(?:mjs|js|ts)(?:\\\\s+--[A-Za-z0-9][A-Za-z0-9_-]*(?:=[A-Za-z0-9_./:@-]+)?)*)$\"\n }\n },\n \"setup_provides_env\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n }\n },\n \"cleanup\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"diagnostics\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation\": {\n \"type\": \"string\"\n },\n \"automation_env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation_env_any\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*(\\\\|[A-Z][A-Z0-9_]*)+$\"\n }\n },\n \"automation_prompt\": {\n \"type\": \"string\"\n },\n \"automation_prompts_json\": {\n \"type\": \"string\"\n },\n \"automation_expected_text\": {\n \"type\": \"string\"\n },\n \"automation_response_timeout_ms\": {\n \"type\": \"string\"\n },\n \"automation_stream_output\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_image_base64_fixture\": {\n \"type\": \"string\"\n },\n \"automation_runner_config_patch_json\": {\n \"type\": \"string\"\n },\n \"automation_restore_runner_config\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_expected_runner_id\": {\n \"type\": \"string\"\n },\n \"automation_reset_debug_chat\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_session_type\": {\n \"type\": \"string\",\n \"enum\": [\"person\", \"group\"]\n },\n \"automation_debug_chat_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_max_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_requests\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_concurrency\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_timeout_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_first_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_max_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_error_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_ok_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_provider_fault_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_expected_prefix\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_prompt_template\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_stream\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_reset\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_fail_on_final_mismatch\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_require_success\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_provider_model_thresholds_json\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_pipeline_name\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_model_name\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fallback_model_names\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_response_text\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_first_token_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_chunk_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_chunk_count\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_first_n\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_every_n\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_models\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fault_status\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_after_first_chunk\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_fake_provider_fail_after_first_chunk_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_after_first_chunk_mode\": {\n \"type\": \"string\",\n \"enum\": [\"disconnect\", \"error_event\"]\n },\n \"automation_fake_provider_fail_after_first_chunk_models\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_dynamic_response\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_filesystem_checks_json\": {\n \"type\": \"string\"\n },\n \"metrics_thresholds_json\": {\n \"type\": \"string\"\n },\n \"load_profile_json\": {\n \"type\": \"string\"\n },\n \"fault_model_json\": {\n \"type\": \"string\"\n },\n \"automation_pipeline_url_env\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n },\n \"automation_pipeline_name_env\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n },\n \"success_patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"failure_patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"expected_failures\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"troubleshooting\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n}\n","skills/schemas/skill-index.schema.json":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/skill-index.schema.json\",\n \"title\": \"LangBot Skills Index\",\n \"type\": \"object\",\n \"required\": [\"generated_by\", \"skills\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"generated_by\": {\n \"type\": \"string\"\n },\n \"skills\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\n \"directory\",\n \"name\",\n \"description\",\n \"references\",\n \"cases\",\n \"case_summaries\",\n \"suites\",\n \"suite_summaries\",\n \"fixtures\",\n \"troubleshooting\",\n \"troubleshooting_summaries\"\n ],\n \"additionalProperties\": false,\n \"properties\": {\n \"directory\": { \"type\": \"string\" },\n \"name\": { \"type\": \"string\" },\n \"description\": { \"type\": \"string\" },\n \"references\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"case_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"mode\", \"area\", \"type\", \"priority\", \"risk\", \"ci_eligible\", \"tags\", \"automation\", \"setup_automation\", \"setup_provides_env\", \"evidence_required\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"mode\": { \"type\": \"string\", \"enum\": [\"agent-browser\", \"probe\"] },\n \"area\": { \"type\": \"string\" },\n \"type\": { \"type\": \"string\" },\n \"priority\": { \"type\": \"string\" },\n \"risk\": { \"type\": \"string\" },\n \"ci_eligible\": { \"type\": \"boolean\" },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation\": { \"type\": \"string\" },\n \"setup_automation\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup_provides_env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"evidence_required\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"suites\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"suite_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"description\", \"type\", \"priority\", \"tags\", \"cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"description\": { \"type\": \"string\" },\n \"type\": { \"type\": \"string\" },\n \"priority\": { \"type\": \"string\" },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"fixtures\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"kind\", \"path\", \"related_cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"kind\": { \"type\": \"string\" },\n \"path\": { \"type\": \"string\" },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"troubleshooting\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"troubleshooting_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"category\", \"related_cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"category\": { \"type\": \"string\" },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n }\n }\n }\n }\n }\n}\n","skills/schemas/suite.schema.json":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/suite.schema.json\",\n \"title\": \"LangBot Skill Test Suite\",\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"description\", \"type\", \"priority\", \"tags\", \"cases\"],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"description\": {\n \"type\": \"string\"\n },\n \"type\": {\n \"type\": \"string\",\n \"enum\": [\n \"smoke\",\n \"regression\",\n \"release_gate\",\n \"exploratory\",\n \"contract\",\n \"performance\",\n \"reliability\",\n \"chaos\",\n \"security\"\n ]\n },\n \"priority\": {\n \"type\": \"string\",\n \"enum\": [\"p0\", \"p1\", \"p2\"]\n },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n }\n }\n}\n","skills/schemas/troubleshooting.schema.json":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/troubleshooting.schema.json\",\n \"title\": \"LangBot Skill Troubleshooting Entry\",\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"symptoms\", \"patterns\", \"likely_causes\", \"fix_steps\", \"verification\"],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"date\": {\n \"type\": \"string\"\n },\n \"category\": {\n \"type\": \"string\",\n \"enum\": [\"product\", \"env_issue\", \"external_dependency\", \"blocked\", \"flaky\"]\n },\n \"symptoms\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"likely_causes\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"fix_steps\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"verification\": {\n \"type\": \"string\"\n },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n}\n","skills/skills.index.json":"{\n \"generated_by\": \"lbs\",\n \"skills\": [\n {\n \"directory\": \"langbot-deploy\",\n \"name\": \"langbot-deploy\",\n \"description\": \"Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \\\"deploy langbot\\\", \\\"langbot docker\\\", \\\"langbot compose\\\", \\\"langbot kubernetes\\\", \\\"langbot config.yaml\\\", \\\"langbot box runtime\\\", \\\"langbot global api key\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-dev\",\n \"name\": \"langbot-dev\",\n \"description\": \"Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \\\"langbot backend\\\", \\\"langbot dev\\\", \\\"langbot api\\\", \\\"add langbot endpoint\\\", \\\"langbot migration\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-eba-adapter-dev\",\n \"name\": \"langbot-eba-adapter-dev\",\n \"description\": \"Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing.\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-env-setup\",\n \"name\": \"langbot-env-setup\",\n \"description\": \"Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths.\",\n \"references\": [\n \"references/browser-access-selection.md\",\n \"references/computer-use.md\",\n \"references/oauth-browser-profile.md\",\n \"references/playwright-mcp.md\",\n \"references/proxy.md\",\n \"references/service-startup.md\",\n \"references/wsl-notes.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-mcp-ops\",\n \"name\": \"langbot-mcp-ops\",\n \"description\": \"Operate a LangBot instance through its built-in MCP (Model Context Protocol) server. Use when an AI agent needs to manage LangBot — list/create/update/delete bots, agents, pipelines, models, knowledge bases, MCP servers, and skills — over MCP instead of raw HTTP. Covers the /mcp endpoint, API-key auth (web-UI lbk_ keys and the config.yaml global key), the tool surface, and client configuration. Triggers on \\\"langbot mcp\\\", \\\"manage langbot via mcp\\\", \\\"langbot /mcp\\\", \\\"langbot mcp server\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-plugin-dev\",\n \"name\": \"langbot-plugin-dev\",\n \"description\": \"Develop, debug, and test LangBot plugins. Use when creating new LangBot plugins, fixing plugin bugs, setting up a LangBot test environment, or testing plugins via WebSocket. Covers plugin component architecture (EventListener, Command, Tool), the plugin SDK API (invoke_llm, get_llm_models, send_message, plugin storage), common pitfalls, and automated WebSocket-based testing. Triggers on \\\"langbot plugin\\\", \\\"lbp\\\", \\\"GroupChatSummary\\\", \\\"plugin debug\\\", \\\"langbot test\\\".\",\n \"references\": [\n \"references/test-env-setup.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-skills-maintenance\",\n \"name\": \"langbot-skills-maintenance\",\n \"description\": \"Maintain the langbot-skills repository with low duplication. Use when adding, editing, or auditing LangBot skills, references, cases, troubleshooting entries, indexes, or periodic entropy-control checks for this skills repository.\",\n \"references\": [\n \"references/curation-workflow.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-space-ops\",\n \"name\": \"langbot-space-ops\",\n \"description\": \"Browse and search the LangBot Space marketplaces (plugins, MCP servers, skills) through the Space MCP server. Use when an AI agent needs to discover LangBot extensions on space.langbot.app over MCP. Covers the /mcp endpoint, Personal Access Token (PAT) auth, the tool surface, and client configuration. Triggers on \\\"langbot space mcp\\\", \\\"search langbot plugins\\\", \\\"langbot marketplace mcp\\\", \\\"space.langbot.app mcp\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-testing\",\n \"name\": \"langbot-testing\",\n \"description\": \"Test LangBot WebUI and core product flows with an automated browser and backend logs. Use when validating the configured LangBot frontend, pipeline Debug Chat, model provider setup and test buttons, bot and knowledge-base UI flows, or troubleshooting failed LangBot end-to-end tests.\",\n \"references\": [\n \"references/agent-runner-qa-workflow.md\",\n \"references/agent-runner-release-gate.md\",\n \"references/dify-agent-runner.md\",\n \"references/langrag-knowledge-base.md\",\n \"references/local-agent-runner-coverage.md\",\n \"references/local-agent-runner.md\",\n \"references/mcp-stdio-testing.md\",\n \"references/model-provider-testing.md\",\n \"references/performance-reliability-testing.md\",\n \"references/pipeline-debug-chat.md\",\n \"references/plugin-e2e-smoke.md\",\n \"references/sandbox-skill-authoring.md\",\n \"references/skill-all-tool-acceptance.md\",\n \"references/troubleshooting.md\",\n \"references/web-ui-testing.md\",\n \"references/workspace-release-testing.md\"\n ],\n \"cases\": [\n \"acp-agent-runner-debug-chat\",\n \"agent-run-ledger-audit\",\n \"agent-runner-async-db-readiness\",\n \"agent-runner-behavior-matrix\",\n \"agent-runner-fixture-contract\",\n \"agent-runner-health-visibility\",\n \"agent-runner-ledger-concurrency\",\n \"agent-runner-ledger-contention\",\n \"agent-runner-ledger-invariants\",\n \"agent-runner-ledger-stress\",\n \"agent-runner-live-install\",\n \"agent-runner-qa-debug-chat\",\n \"agent-runner-release-preflight\",\n \"agent-runner-runtime-chaos\",\n \"bot-event-routing-product-flow\",\n \"box-mcp-heartbeat-recovery\",\n \"claude-code-agent-debug-chat\",\n \"codex-agent-debug-chat\",\n \"dify-agent-debug-chat\",\n \"langbot-fake-provider-debug-chat-cross-pipeline-isolation\",\n \"langbot-fake-provider-debug-chat-fault-recovery\",\n \"langbot-fake-provider-debug-chat-load\",\n \"langbot-fake-provider-debug-chat-slow-load\",\n \"langbot-fault-taxonomy-contract\",\n \"langbot-live-backend-latency\",\n \"langbot-live-backend-log-health\",\n \"langbot-live-control-plane-api\",\n \"langbot-overhead-accounting-contract\",\n \"langbot-space-debug-chat-concurrency-smoke\",\n \"langrag-kb-retrieve\",\n \"langrag-parser-golden-e2e\",\n \"langrag-sentinel-kb-discover\",\n \"local-agent-basic-debug-chat\",\n \"local-agent-combo-rag-compaction-tool-debug-chat\",\n \"local-agent-complex-coding-task-debug-chat\",\n \"local-agent-context-compaction-debug-chat\",\n \"local-agent-effective-prompt-debug-chat\",\n \"local-agent-model-fallback-before-first-chunk-debug-chat\",\n \"local-agent-multimodal-debug-chat\",\n \"local-agent-multitool-rag-compaction-debug-chat\",\n \"local-agent-nonstreaming-debug-chat\",\n \"local-agent-parallel-tools-rag-compaction-debug-chat\",\n \"local-agent-plugin-tool-call-debug-chat\",\n \"local-agent-rag-debug-chat\",\n \"local-agent-rag-multimodal-debug-chat\",\n \"local-agent-steering-debug-chat\",\n \"local-agent-streaming-post-commit-failure-debug-chat\",\n \"local-agent-tool-error-recovery-debug-chat\",\n \"local-agent-tool-loop-limit-debug-chat\",\n \"mcp-stdio-register\",\n \"mcp-stdio-tool-call\",\n \"pipeline-debug-chat\",\n \"pipeline-debug-chat-performance\",\n \"plugin-e2e-smoke\",\n \"provider-deepseek\",\n \"qa-plugin-smoke-live-install\",\n \"sandbox-skill-authoring-e2e\",\n \"sandbox-skill-authoring-edit-existing-e2e\",\n \"skill-discovery-via-mcp-gateway\",\n \"webui-login-state\",\n \"wizard-onebot-agent-runtime\",\n \"wizard-runner-marketplace-catalog\",\n \"wizard-scenario-routing\",\n \"workspace-compatibility-preflight\",\n \"workspace-plugin-pages-smoke\",\n \"workspace-repository-contracts\"\n ],\n \"case_summaries\": [\n {\n \"id\": \"acp-agent-runner-debug-chat\",\n \"title\": \"ACP Runner can answer through Debug Chat using real remote Claude\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"acp\",\n \"claude\",\n \"external-runner\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_ACP_RUNNER_PIPELINE_URL\",\n \"LANGBOT_ACP_RUNNER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"agent-run-ledger-audit\",\n \"title\": \"Persisted Runner run ledger passes end-to-end invariants\",\n \"mode\": \"probe\",\n \"area\": \"agent\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"ledger\",\n \"audit\",\n \"tools\"\n ],\n \"automation\": \"scripts/e2e/agent-run-ledger-audit.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\",\n \"metrics\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-async-db-readiness\",\n \"title\": \"Runner async DB readiness probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"async-db\",\n \"aiosqlite\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-async-db-readiness.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-behavior-matrix\",\n \"title\": \"Runner deterministic behavior matrix probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"deterministic-runner\",\n \"protocol\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-behavior-matrix.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-fixture-contract\",\n \"title\": \"QA Runner fixture contract probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"fixture\",\n \"deterministic-runner\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-fixture-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-health-visibility\",\n \"title\": \"Agent configuration shows whether its selected runner is usable\",\n \"mode\": \"agent-browser\",\n \"area\": \"agent\",\n \"type\": \"feature\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"agent\",\n \"productization\",\n \"release-gate\"\n ],\n \"automation\": \"scripts/e2e/agent-runner-health-visibility.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-concurrency\",\n \"title\": \"Runner run ledger concurrency and auth pytest probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"pytest\",\n \"run-ledger\",\n \"concurrency\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-concurrency.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-contention\",\n \"title\": \"Runner ledger SQLite contention probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"stress\",\n \"ledger\",\n \"concurrency\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-contention.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-invariants\",\n \"title\": \"Runner ledger schema and status invariants probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"ledger\",\n \"invariant\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-invariants.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-stress\",\n \"title\": \"Runner ledger lightweight stress probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"stress\",\n \"ledger\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-stress.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-live-install\",\n \"title\": \"QA Runner package installs and registers in LangBot\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"plugin\",\n \"local-install\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/install-qa-plugin-smoke.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-qa-debug-chat\",\n \"title\": \"QA Runner returns deterministic output through Debug Chat\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"pipeline\",\n \"debug-chat\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"case:agent-runner-live-install\",\n \"node:scripts/e2e/ensure-qa-agent-runner-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_QA_RUNNER_PIPELINE_URL\",\n \"LANGBOT_QA_RUNNER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-release-preflight\",\n \"title\": \"Agent runner release gate preflight validates environment readiness\",\n \"mode\": \"agent-browser\",\n \"area\": \"release\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"release-gate\",\n \"preflight\",\n \"environment\"\n ],\n \"automation\": \"scripts/e2e/agent-runner-release-preflight.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-runtime-chaos\",\n \"title\": \"Runner SDK runtime chaos pytest probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"pytest\",\n \"runtime\",\n \"sdk\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-runtime-chaos.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"bot-event-routing-product-flow\",\n \"title\": \"Bot event routing and adapter input can be inspected from the WebUI\",\n \"mode\": \"agent-browser\",\n \"area\": \"bot\",\n \"type\": \"feature\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"bot\",\n \"eba\",\n \"event-routing\",\n \"productization\"\n ],\n \"automation\": \"scripts/e2e/bot-event-routing-product-flow.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"box-mcp-heartbeat-recovery\",\n \"title\": \"Box heartbeat restores an existing stdio MCP tool session\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"box\",\n \"mcp\",\n \"stdio\",\n \"reliability\",\n \"chaos\",\n \"fault-injection\"\n ],\n \"automation\": \"skills/langbot-testing/probes/box-mcp-heartbeat-recovery.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"metrics\",\n \"api_diagnostic\",\n \"resource_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"claude-code-agent-debug-chat\",\n \"title\": \"Claude Code Agent can retrieve LangBot history through MCP\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"claude-code\",\n \"external-runner\",\n \"mcp\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"codex-agent-debug-chat\",\n \"title\": \"Codex Agent can retrieve LangBot history through MCP\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"codex\",\n \"external-runner\",\n \"mcp\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"dify-agent-debug-chat\",\n \"title\": \"Dify Runner returns a response through Pipeline Debug Chat\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"provider\",\n \"priority\": \"p2\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"dify\",\n \"agent-runner\",\n \"pipeline\"\n ],\n \"automation\": \"\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-cross-pipeline-isolation\",\n \"title\": \"LangBot Debug Chat fake-provider cross-pipeline isolation probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"reliability\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"isolation\",\n \"concurrency\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-cross-pipeline-isolation.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-cross-pipelines.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_A_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_A_NAME\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_B_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_B_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-fault-recovery\",\n \"title\": \"LangBot Debug Chat fake-provider fault recovery probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"chaos\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"fault-injection\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-load\",\n \"title\": \"LangBot Debug Chat controlled fake-provider load probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"load\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-slow-load\",\n \"title\": \"LangBot Debug Chat slow fake-provider load probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"slow-provider\",\n \"load\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fault-taxonomy-contract\",\n \"title\": \"LangBot fault taxonomy and cleanup contract\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": true,\n \"tags\": [\n \"reliability\",\n \"chaos\",\n \"contract\",\n \"synthetic\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-fault-taxonomy-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-backend-latency\",\n \"title\": \"LangBot live backend basic latency probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"live-backend\",\n \"latency\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-backend-latency.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-backend-log-health\",\n \"title\": \"LangBot live backend log health probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"reliability\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"live-backend\",\n \"backend-log\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-backend-log-health.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"backend_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-control-plane-api\",\n \"title\": \"LangBot live control-plane API probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"reliability\",\n \"live-backend\",\n \"control-plane\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-control-plane-api.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-overhead-accounting-contract\",\n \"title\": \"LangBot overhead accounting metrics contract\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": true,\n \"tags\": [\n \"performance\",\n \"metrics\",\n \"contract\",\n \"synthetic\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-overhead-accounting-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"resource_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-space-debug-chat-concurrency-smoke\",\n \"title\": \"LangBot Debug Chat real Space-provider concurrency smoke\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"space\",\n \"live-provider\",\n \"smoke\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_PIPELINE_URL\",\n \"LANGBOT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_MODEL_UUID\",\n \"LANGBOT_E2E_MODEL_UUID\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langrag-kb-retrieve\",\n \"title\": \"LangRAG knowledge base ingests and retrieves a sentinel document\",\n \"mode\": \"agent-browser\",\n \"area\": \"knowledge\",\n \"type\": \"feature\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"langrag\",\n \"knowledge\",\n \"rag\"\n ],\n \"automation\": \"scripts/e2e/langrag-kb-retrieve.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"langrag-parser-golden-e2e\",\n \"title\": \"LangRAG and GeneralParsers retrieve a structured golden document\",\n \"mode\": \"agent-browser\",\n \"area\": \"knowledge\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"golden\",\n \"e2e\",\n \"langrag\",\n \"parser\",\n \"knowledge\"\n ],\n \"automation\": \"\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"langrag-sentinel-kb-discover\",\n \"title\": \"Existing LangRAG sentinel knowledge base is discoverable\",\n \"mode\": \"probe\",\n \"area\": \"knowledge\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"langrag\",\n \"knowledge\",\n \"rag\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/ensure-langrag-sentinel-kb.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-basic-debug-chat\",\n \"title\": \"Local Agent Debug Chat returns a deterministic streaming response\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"streaming\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-combo-rag-compaction-tool-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and plugin tool result in one Debug Chat run\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-complex-coding-task-debug-chat\",\n \"title\": \"Local Agent completes a multi-file coding task with iterative verification\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"agentic\",\n \"coding\",\n \"sandbox\",\n \"tools\",\n \"long-running\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/reset-complex-agent-task.mjs\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_COMPLEX_AGENT_WORKSPACE\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\",\n \"metrics\"\n ]\n },\n {\n \"id\": \"local-agent-context-compaction-debug-chat\",\n \"title\": \"Local Agent compacts long Debug Chat history and preserves older facts\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"context\",\n \"compaction\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-effective-prompt-debug-chat\",\n \"title\": \"Local Agent consumes host effective prompt after PromptPreProcessing\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"prompt\",\n \"plugin\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-model-fallback-before-first-chunk-debug-chat\",\n \"title\": \"Local Agent falls back when the primary model fails before streaming starts\",\n \"mode\": \"probe\",\n \"area\": \"pipeline\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"fallback\",\n \"streaming\",\n \"fake-provider\",\n \"fault-injection\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"local-agent-multimodal-debug-chat\",\n \"title\": \"Local Agent Debug Chat preserves uploaded image input\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"multimodal\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-multitool-rag-compaction-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and two-step plugin tool loop\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"tool-loop\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-nonstreaming-debug-chat\",\n \"title\": \"Local Agent Debug Chat returns a deterministic response with UI streaming disabled\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"non-streaming\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-parallel-tools-rag-compaction-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and parallel plugin tools\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"parallel-tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-plugin-tool-call-debug-chat\",\n \"title\": \"Local Agent can call a plugin-provided tool\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"plugin\",\n \"tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-rag-debug-chat\",\n \"title\": \"Local Agent Debug Chat answers from a LangRAG knowledge base\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-rag-multimodal-debug-chat\",\n \"title\": \"Local Agent preserves image input while using RAG context\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"rag\",\n \n\n","skills/skills/langbot-deploy/SKILL.md":"---\nname: langbot-deploy\ndescription: Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \"deploy langbot\", \"langbot docker\", \"langbot compose\", \"langbot kubernetes\", \"langbot config.yaml\", \"langbot box runtime\", \"langbot global api key\".\n---\n\n# LangBot Deployment & Configuration\n\nCovers running LangBot in production. For development see `langbot-dev`.\n\nRead https://langbot.app/llms.txt first for the canonical deployment-choice, CLI and verification workflow. The summary below applies when selecting hosting.\n\n## Deployment choice and optional managed hosting\n\nRespect the user's deployment preference and reuse an existing instance when available. Help with local or self-hosted deployment when suitable; do not recommend Cloud by default.\n\nOffer [LangBot Cloud](https://space.langbot.app/cloud) when the user wants managed hosting without maintaining a server, the local environment is unsuitable, or a deployment problem remains difficult after reasonable troubleshooting. Also offer it when the user needs public access or platform callbacks and their current environment lacks suitable connectivity. Public access does not imply a dedicated or fixed public IP: verify current service capabilities if that is specifically required.\n\nIf the user wants to support LangBot's continued development and open-source maintenance, mention the official hosted service as an optional way to do so, without pressure or repeated appeals. Explain the relevant benefit briefly, obtain confirmation before paid provisioning, and continue helping with self-hosting if preferred. Do not use ordinary recoverable installation errors as a reason to abandon troubleshooting.\n\n## Docker Compose (recommended)\n\n```bash\ngit clone https://github.com/langbot-app/LangBot\ncd LangBot/docker\n\n# Full stack (sandbox/Box + stdio MCP hosting + skill add/edit enabled)\ndocker compose --profile all up\n\n# Basic (no Box runtime)\ndocker compose up\n```\n\nThe `all` / `box` profile starts three services:\n\n- `langbot` — main app, serves API + UI on `:5300`.\n- `langbot_plugin_runtime` — plugin runtime (control `:5400`, debug `:5401`).\n- `langbot_box` — Box sandbox runtime (`:5410`). Uses the host Docker socket to\n spawn sandbox containers, so the **Box root host path and in-container path\n must be identical** (`BOX__LOCAL__HOST_ROOT=${LANGBOT_BOX_ROOT:-${PWD}/data/box}`).\n OSS allows its RPC and managed-process relay to run without a token when both\n sides leave `LANGBOT_BOX_CONTROL_TOKEN` unset. For an exposed endpoint, set\n the same value of at least 32 non-whitespace characters in both the LangBot\n and Box containers. Generate it once with `openssl rand -hex 32`; never put\n it in `box.runtime.endpoint` or commit it to config.\n\nA Compose deployment may optionally set\n`LANGBOT_PLUGIN_RUNTIME_CONTROL_TOKEN` on both `langbot` and\n`langbot_plugin_runtime` when port 5400 needs shared-secret protection. OSS\ndefaults to leaving it unset on both sides. If enabled, generate one value with\n`openssl rand -hex 32`; configuring only one side causes the control connection\nto fail. Kubernetes may use the `langbot-plugin-runtime-control` Secret shown in\n`docker/kubernetes.yaml`.\n\nWith Box off, the dashboard/skills list stays visible (read-only) but sandbox\ntools, skill add/edit, and stdio MCP are disabled. Set `box.enabled: false`\n(or `BOX__ENABLED=false`) to match.\n\n## Kubernetes\n\nSee `docker/kubernetes.yaml` and the deployment guide at\nhttps://langbot.app/docs. `docker/deploy-k8s-test.sh` is a test helper.\n\n## config.yaml (generated at `data/config.yaml` on first run)\n\nTop-level sections: `api`, `system`, `command`, `concurrency`, `proxy`,\n`database`, `vdb`, `storage`, `plugin`, `monitoring`, `box`, `space`.\n\nKey settings:\n\n| Key | Meaning |\n| --- | --- |\n| `api.port` | HTTP API + UI port (default 5300) |\n| `api.global_api_key` | **Global API key** for the HTTP API + MCP server. Non-empty = accepted with no login/DB record; no `lbk_` prefix required. Empty = disabled. Plaintext — trusted/internal only, serve over HTTPS. |\n| `plugin.runtime_ws_url` | Standalone plugin runtime WS URL (e.g. `ws://langbot_plugin_runtime:5400/control/ws`) |\n| `box.enabled` | Master switch for the Box sandbox runtime |\n| `box.backend` | `local` (Docker/nsjail autopick) / `docker` / `nsjail` / `e2b` / explicit unsafe `host`; env override `BOX__BACKEND` |\n| `box.runtime.endpoint` | External Box runtime URL (e.g. `ws://127.0.0.1:5410`); empty = local auto-managed |\n\nMany keys have `ENV__SUBKEY` overrides (e.g. `BOX__BACKEND`, `BOX__ENABLED`).\n\n## Runtimes & flags\n\n- LangBot started directly spawns the plugin runtime over **stdio**.\n- In containers it connects to a standalone runtime over **WebSocket**; start\n with `--standalone-runtime`.\n- Box has a parallel `--standalone-box` flag; the Docker box host is\n `langbot_box:5410`.\n- `box.backend: host` runs commands directly as the Box Runtime system user.\n It is never auto-selected, provides no sandbox isolation, and is only for\n trusted local development. A WebSocket-controlled host backend requires\n `LANGBOT_BOX_CONTROL_TOKEN`; local stdio control is allowed.\n\n## Global API key — enabling for agents/automation\n\n```yaml\n# data/config.yaml\napi:\n port: 5300\n global_api_key: 'a-strong-secret' # empty disables it\n```\n\nThis key authenticates both the HTTP API and the MCP server (`/mcp`) without a\nlogin session. See `langbot-mcp-ops` for using it, and `docs/API_KEY_AUTH.md`.\n\n## Pitfalls\n\n- \"No supported sandbox backend (Docker / nsjail / E2B)\" with Docker running\n usually means the user isn't in the `docker` group →\n `sudo usermod -aG docker ` and restart in a new shell.\n- Do not use `box.backend: host` as a production fallback. It cannot enforce\n image, filesystem, network, PID, CPU, memory, or storage isolation.\n- Box root host/container path mismatch breaks sandbox container creation.\n- Don't commit a non-empty `api.global_api_key` to version control.\n","skills/skills/langbot-dev/SKILL.md":"---\nname: langbot-dev\ndescription: Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \"langbot backend\", \"langbot dev\", \"langbot api\", \"add langbot endpoint\", \"langbot migration\".\n---\n\n# LangBot Core Development\n\nThis skill covers developing the LangBot core (the main repo), distinct from\nplugin development (see `langbot-plugin-dev`) and deployment (`langbot-deploy`).\n\n## Stack\n\n- **Backend**: Python `>=3.11,<4.0`, deps via `uv`. Framework: **Quart** (async\n Flask). Serves the HTTP API + pre-built web UI on `http://127.0.0.1:5300`.\n- **Frontend** (`web/`): **Vite + React Router 7 + shadcn/ui + Tailwind**,\n managed by `pnpm`. Dev server on `:3000`. (NOT Next.js — `dev` script is `vite`.)\n\n## Dev environment\n\n```bash\n# Backend\npip install uv\nuv sync --dev\nuv run main.py # API + UI on http://127.0.0.1:5300\n\n# Frontend (separate terminal)\ncd web\ncp .env.example .env\npnpm install\npnpm dev # http://127.0.0.1:3000 (reads VITE_API_BASE_URL)\n\n# Lint/format hooks (CI runs the same checks)\nuv run pre-commit install\n```\n\nFirst run generates `data/config.yaml`; DB defaults to SQLite (PostgreSQL\nsupported). Migrations run automatically on startup.\n\n## Repo layout (key paths)\n\n```\nsrc/langbot/\n├── __main__.py # entrypoint, CLI flags (--standalone-runtime/-box/--debug)\n├── pkg/\n│ ├── api/\n│ │ ├── http/ # Quart controllers + services\n│ │ │ ├── controller/groups/ # route groups (@group.group_class)\n│ │ │ └── service/ # business logic (called by controllers AND MCP)\n│ │ └── mcp/ # MCP server (server.py = tools, mount.py = ASGI dispatch)\n│ ├── core/ # app bootstrap, stages, task manager\n│ ├── platform/ provider/ pipeline/ plugin/ box/ skill/ rag/ vector/\n│ ├── command/ persistence/ storage/ config/ entity/ telemetry/\n│ └── templates/config.yaml # config template (top-level: api, system, plugin, box, space...)\n├── web/ # Vite SPA\n└── docker/ # compose deployment\n```\n\n## HTTP API auth model\n\nRoute auth is declared per-route via `AuthType` in\n`pkg/api/http/controller/group.py`:\n\n- `NONE` — public.\n- `USER_TOKEN` — web UI JWT (`Authorization: Bearer `).\n- `API_KEY` — `X-API-Key` or `Authorization: Bearer `.\n- `USER_TOKEN_OR_API_KEY` — either.\n\nAuthenticated routes receive an immutable `RequestContext` containing the\nprincipal, authorized Workspace membership, fixed-role permissions, instance,\nrequest id, and placement generation. A browser's `X-Workspace-Id` is only a\nselector and is always checked against the Account membership. Tenant services\nmust accept this context (or an explicit trusted execution context) and fail\nclosed when it is absent.\n\nAPI-key authentication accepts:\n\n1. the **global key** from `config.yaml` `api.global_api_key` only for a\n community instance with exactly one local Workspace, then\n2. **web-UI keys** whose one-time `lbk_` secret is stored only as a hash and is\n bound to one Workspace, explicit scopes, status, and optional expiry.\n\nAn API key derives its Workspace from the key record and ignores a caller's\nWorkspace selector. Public Bot/Webhook routes similarly derive Workspace from\nthe opaque owning resource rather than a header.\n\nRoute groups self-register via `@group.group_class(name, path)` and are\ndiscovered by `importutil.import_modules_in_pkg`.\n\n## Adding an API endpoint\n\n1. Add/extend a controller in `pkg/api/http/controller/groups/` and the matching\n service method in `pkg/api/http/service/`.\n2. Pick the right `AuthType`.\n3. **If the endpoint should be agent-accessible, add/adjust the matching MCP tool\n in `pkg/api/mcp/server.py` and update the `langbot-mcp-ops` skill.** API and\n MCP surface must stay aligned (see `AGENTS.md`).\n4. If `lbctl` uses the route or capability, check the separate\n [`langbot-cli`](https://github.com/langbot-app/langbot-cli) client for compatibility.\n5. Update `docs/service-api-openapi.json` if you maintain the OpenAPI overview.\n\n## Database migrations (Alembic)\n\nSingle migration set supports SQLite + PostgreSQL. Files in\n`src/langbot/pkg/persistence/alembic/versions/`.\n\n```bash\n# From project root (needs data/config.yaml)\nuv run python -m langbot.pkg.persistence.alembic_runner autogenerate \"description\"\n```\n\n## Standards\n\n- All code comments/docstrings in **English**; user-facing strings need **i18n**\n (`en_US` + `zh_Hans` minimum, `ja_JP` where present).\n- Consider toC and toB compatibility + security.\n- Commit format: `(): ` (feat/fix/docs/refactor/...).\n\n## Tests\n\n```bash\nuv run pytest tests/unit_tests -q # unit tests\nuv run pytest tests/unit_tests/api -q # API service tests\nuv run python tests/manual/mcp_smoke.py # MCP server e2e smoke\n```\n\n## See also\n\n- `langbot-plugin-dev` — plugin SDK / runtime development.\n- `langbot-testing` — WebUI/e2e QA harness (`bin/lbs`).\n- `langbot-deploy` — Docker/compose deployment + config.\n- `langbot-mcp-ops` — operating the LangBot MCP server.\n- [`langbot-cli`](https://github.com/langbot-app/langbot-cli) — `lbctl`, a standalone Service API client for managing running Workspaces.\n","skills/skills/langbot-eba-adapter-dev/SKILL.md":"---\nname: langbot-eba-adapter-dev\ndescription: Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing.\n---\n\n# LangBot EBA Adapter Development\n\nUse this skill when implementing or reviewing a LangBot platform adapter under the Event-Based Agents architecture.\n\n## Controlling a running instance via MCP\n\nBeyond writing code, you can **drive a live LangBot instance over MCP** — no raw\nHTTP needed. Two MCP servers exist (both reuse existing API keys; see `AGENTS.md`):\n\n- **LangBot instance** — `http://:5300/mcp` (auth: web-UI `lbk_` key or the\n `api.global_api_key` from `config.yaml`). Manage bots, pipelines, models,\n knowledge bases, and skills. See the **`langbot-mcp-ops`** skill.\n- **LangBot Space marketplace** — `https://space.langbot.app/mcp` (auth: Personal\n Access Token). Search plugins / MCP servers / skills. See the\n **`langbot-space-ops`** skill.\n\n> Any change to an agent-accessible HTTP API endpoint must keep the matching MCP\n> tool and these skills in sync.\n\n## Core Rule\n\nDo not let platform-native event or message shapes leak into LangBot's common path. Each adapter must convert incoming SDK objects into unified EBA entities before dispatch:\n\n- Events: `langbot_plugin.api.entities.builtin.platform.events`\n- Message chains: `langbot_plugin.api.entities.builtin.platform.message.MessageChain`\n- Users/groups/members: `langbot_plugin.api.entities.builtin.platform.entities`\n- Raw platform objects may remain only in `source_platform_object` for debugging or platform-specific escape hatches.\n\n## Start Here\n\n1. Read the EBA design docs in `LangBot/docs/event-based-agents/`.\n2. Read the architecture-level acceptance checklist before writing or validating code:\n - `LangBot/docs/event-based-agents/adapters/acceptance-checklist.md`\n3. Read the current reference adapter before writing code. Prefer Telegram first:\n - `LangBot/src/langbot/pkg/platform/adapters/telegram/`\n - `LangBot/docs/event-based-agents/adapters/telegram.md`\n4. Read the legacy source adapter for the target platform:\n - `LangBot/src/langbot/pkg/platform/sources/.py`\n - `LangBot/src/langbot/pkg/platform/sources/.yaml`\n5. Inspect SDK entity definitions in `langbot-plugin-sdk/src/langbot_plugin/api/entities/builtin/platform/`.\n6. Search before assuming APIs. Platform SDKs change often.\n\n## Adapter Layout\n\nCreate one directory per adapter:\n\n```text\nLangBot/src/langbot/pkg/platform/adapters//\n├── __init__.py\n├── adapter.py\n├── api_impl.py\n├── event_converter.py\n├── manifest.yaml\n├── message_converter.py\n├── platform_api.py\n├── types.py\n└── .svg\n```\n\nAdd optional helpers such as `voice.py` only when the platform has a real domain-specific surface.\n\nEnsure `pyproject.toml` package data includes adapter assets:\n\n```toml\npackage-data = { \"langbot\" = [\"templates/**\", \"pkg/platform/sources/*\", \"pkg/platform/adapters/**\", ...] }\n```\n\n## Implementation Checklist\n\n- `manifest.yaml` declares `metadata.name`, config schema, supported events, common APIs, and platform-specific APIs.\n- `adapter.py` creates the platform client, subscribes to native events, filters self/bot loops where appropriate, calls `event_converter.target2yiri(...)`, then dispatches the EBA event.\n- `event_converter.py` maps native events to EBA event classes such as `MessageReceivedEvent`, `MessageEditedEvent`, `MessageDeletedEvent`, `MessageReactionEvent`, `MemberJoinedEvent`, `BotInvitedToGroupEvent`, and `PlatformSpecificEvent`.\n- `message_converter.py` maps native messages to `MessageChain`, and maps `MessageChain` back to the platform send format.\n- `api_impl.py` implements common EBA APIs: send, reply, edit, delete, forward, user/group/member lookup, moderation, upload/file URL, leave group.\n- `platform_api.py` keeps platform-specific calls behind `call_platform_api(action, params)`.\n- Unsupported common APIs must raise explicit SDK platform errors such as `NotSupportedError`; do not silently no-op.\n- Destructive APIs such as kick, ban, leave, delete, or moderation must be gated in live tests and documented.\n\n## Conversion Contract\n\nFor message events, the common shape should look like this regardless of platform:\n\n```python\nplatform_events.MessageReceivedEvent(\n type=\"message.received\",\n adapter_name=\"\",\n message_id=,\n message_chain=platform_message.MessageChain([...]),\n sender=platform_entities.User(...),\n chat_type=platform_entities.ChatType.PRIVATE or ChatType.GROUP,\n chat_id=,\n group=platform_entities.UserGroup(...) or None,\n source_platform_object=,\n)\n```\n\nMessage content should use common components:\n\n- `Source` for original message id/time when available.\n- `Plain` for text.\n- `At` / `AtAll` for mentions.\n- `Image`, `Voice`, `File` for media.\n- `Forward` only when the platform can represent or emulate it safely.\n\nIf a platform event cannot cleanly map to a common event, emit `PlatformSpecificEvent` with a compact `action` and structured `data`.\n\n## Unit Tests\n\nAdd focused tests under `LangBot/tests/unit_tests/platform/test__eba_adapter.py`.\n\nCover at least:\n\n- Manifest supported events match adapter `supported_events()`.\n- Manifest supported APIs match adapter `supported_apis()`.\n- Platform API map matches manifest actions.\n- Dispatcher chooses the most specific EBA listener.\n- Message converter maps every supported common component both directions where possible:\n - `Source`\n - `Plain`\n - `At`\n - `AtAll`\n - `Image`\n - `Voice`\n - `File`\n - `Quote`\n - `Face`\n - `Forward`\n - `Unknown`\n - mixed chains preserving order\n- Event converter maps message received/edited/deleted/reaction, raw uncached gateway events, member events, and bot join/leave events.\n- Send/reply methods pass correct platform kwargs and return `MessageResult`.\n\nRun the existing reference adapter tests too:\n\n```bash\ncd LangBot\nuv run pytest tests/unit_tests/platform/test__eba_adapter.py tests/unit_tests/platform/test_telegram_eba_adapter.py\nuv run python -m py_compile tests/e2e/live__eba_probe.py\ngit diff --check\n```\n\n## Live Test Workflow\n\nDirect adapter live probes are useful diagnostics, but they are not sufficient acceptance evidence for EBA. Treat `tests/e2e/live__eba_probe.py` as an auxiliary tool only. The final adapter record must distinguish:\n\n- `plugin-e2e-ui`: real SDK plugin through standalone runtime, LangBot core, adapter, and a real/simulator UI action. This can mark an inbound UI item complete.\n- `plugin-e2e-protocol`: real SDK plugin through standalone runtime, LangBot core, adapter, and a protocol-boundary injected event. This is useful evidence but must not be claimed as UI coverage.\n- `plugin-e2e-outbound`: real SDK plugin calls an API and the bot output is visible in the real/simulator UI. This can mark send/API coverage complete.\n- `adapter-live`: direct adapter probe connected to a real/simulator endpoint. This is auxiliary only.\n- `unit`: mocked conversion/API-shape coverage. This is auxiliary only.\n- `not-supported`: platform protocol or SDK has no equivalent. Must include the reason.\n- `blocked`: intended capability could not be verified. This is not complete.\n\nWrite a live probe in `LangBot/tests/e2e/live__eba_probe.py`. It should:\n\n1. Read token/client ids from environment variables or CLI args.\n2. Start the adapter directly.\n3. Register an EBA listener and write JSONL evidence to `LangBot/data/temp/`.\n4. Wait for a real user/platform event instead of fabricating the entrypoint.\n5. Exercise common APIs and `call_platform_api` actions.\n6. Observe returned gateway events for edit/delete/reaction/member/bot lifecycle where available.\n7. Print a summary containing passed, failed, skipped, and observed event types.\n8. Redact or avoid printing secrets.\n9. Keep destructive operations behind flags and run them last.\n\nUse Computer Use when the user asks for real platform end-to-end coverage. Actually send messages/click reactions in the platform UI or otherwise trigger real user-side events; do not replace that with unit tests.\n\nFor media/component acceptance, keep the direction and trigger source explicit:\n\n- Real inbound media only counts when a human-side platform UI or simulator UI sends the image/file/voice to the bot and the plugin JSONL records the corresponding common component.\n- Bot outbound media only proves `send_message`/adapter send conversion. It does not prove inbound conversion.\n- Protocol-boundary injection, such as sending a OneBot event directly into a reverse WebSocket adapter, is useful and should be labelled `plugin-e2e-protocol`, but it must not be reported as UI-level end-to-end media upload.\n- If the UI cannot send or upload the media, record the item as `blocked` with the exact client/simulator limitation.\n\n## Standalone Runtime + Plugin Test\n\nWhen validating the whole LangBot EBA path, test with the SDK standalone runtime and a real test plugin. This is the required acceptance path; direct adapter calls do not prove the EBA architecture path.\n\nThe required path is:\n\n```text\nReal platform / simulator UI\n -> platform SDK native event\n -> adapter event converter\n -> unified EBA event/entity/message types\n -> LangBot core event dispatch\n -> standalone SDK runtime\n -> real test plugin listener\n -> plugin calls platform APIs through SDK\n -> LangBot core API dispatch\n -> adapter API implementation\n -> real platform / simulator UI\n```\n\nTypical shape:\n\n```bash\n# Terminal 1, SDK repo\ncd langbot-plugin-sdk\nuv run python -m langbot_plugin.cli.__init__ rt \\\n --debug-only \\\n --ws-control-port 5400 \\\n --ws-debug-port 5401 \\\n --skip-deps-check\n\n# Terminal 2, LangBot repo\ncd LangBot\nexport PYTHONPATH=/absolute/path/to/langbot-plugin-sdk/src:${PYTHONPATH:-}\nuv run main.py --standalone-runtime\n\n# Terminal 3, plugin directory\nexport DEBUG_RUNTIME_WS_URL=ws://127.0.0.1:5401/plugin/ws\nexport EBA_PROBE_LOG=/absolute/path/to/LangBot/data/temp/_eba_plugin_probe.jsonl\nexport EBA_PROBE_API=1\nexport EBA_PROBE_COMPONENT_SWEEP=1\nexport EBA_PROBE_PLATFORM_API=1\nuv --project /absolute/path/to/langbot-plugin-sdk run python -m langbot_plugin.cli.__init__ run\n```\n\nUse an EBA probe plugin that subscribes to all relevant EBA event classes and runs SDK API calls after the first `MessageReceived`.\n\nThe plugin evidence should be JSONL and include:\n\n- event class and `event.type`\n- adapter name\n- chat type and chat ID\n- sender/user/group IDs with secrets redacted\n- `bot_uuid` and `adapter_name`, proving LangBot filled common routing fields before plugin dispatch\n- received `message_chain` component list\n- API action name, input summary, result or error\n- unsupported or blocked reason when an item is skipped\n\nFor full adapter acceptance, enable both probe sweeps:\n\n- `EBA_PROBE_COMPONENT_SWEEP=1` sends the required outbound message components through `send_message`.\n- `EBA_PROBE_PLATFORM_API=1` calls common safe APIs plus selected `call_platform_api` actions for the adapter.\n\nThe SDK must support `plugin.call_platform_api(bot_uuid, action, params)` for platform-specific acceptance. If the SDK cannot call a platform-specific action from the plugin, the adapter cannot be fully accepted even if direct adapter probes pass.\n\n## Required EBA Acceptance Coverage\n\nBefore marking an adapter migrated, fill out an adapter record against `LangBot/docs/event-based-agents/adapters/acceptance-checklist.md`.\n\nAt minimum, the record must cover these categories:\n\n- Message receive component tests through `plugin-e2e-ui`: `Source`, `Plain`, `At`, `AtAll`, `Image`, `Voice`, `File`, `Quote`, `Face`, `Forward`, `Unknown`, and mixed chains where the platform supports them. Protocol-only receive evidence must be labelled `plugin-e2e-protocol`.\n- Message send component tests through `plugin-e2e-outbound`: `Plain`, `At`, `AtAll`, `Image`, `Voice`, `File`, `Quote`, `Face`, `Forward`, and mixed chains where the platform supports them.\n- Every event declared in `manifest.yaml -> spec.supported_events`.\n- Every common API declared in `manifest.yaml -> spec.supported_apis.required` and `optional`.\n- Every action declared in `manifest.yaml -> spec.platform_specific_apis`.\n- Compatibility tests for manifest declarations, legacy message listener fallback, EBA listener specificity, bot self-message filtering, and `source_platform_object` reply/debug behavior.\n\nDo not declare an event or API in the manifest unless it has an implementation path and an acceptance entry. If a platform or simulator lacks a capability, document it as `not-supported` or `blocked` rather than silently omitting the test.\n\n## Common Pitfalls\n\n- `get_bots()` may return bot dictionaries, not UUID strings. Probe plugins should select an enabled dict and pass `bot[\"uuid\"]` to `get_bot_info()` and `send_message()`.\n- Make sure the probe subscribes to every event you claim to verify. Missing `MessageDeleted` subscription can make a working adapter look untested.\n- Some platforms emit both cached and raw gateway events, producing duplicate evidence for delete/reaction. Count this explicitly; do not treat duplicates as failure unless semantics differ.\n- Self-message filtering is platform-specific. Filter bot-originated `message.received` loops, but do not accidentally filter edit/delete events needed for bot-owned API probes.\n- Reaction events may be filtered for bot self reactions. To test user reaction add/remove, use real UI interaction or a real user token path if permitted.\n- File uploads usually happen as message attachments. A standalone `upload_file` API may need to be `NotSupportedError`.\n- Live probes should not leak bot tokens through command output, logs, docs, or final answers.\n- Discord requires privileged intents for message content and members. Missing intents can look like converter bugs.\n- Telegram Bot API exposes only limited member lists; document capability gaps.\n- Do not mark moderation APIs verified unless they ran against a disposable target member/bot.\n- If `leave_group` is tested, run it last because the test bot will be removed from the server/group.\n- Restore local LangBot DB/test state after live runs if you enabled temporary bots or changed plugin settings.\n\n## Documentation Record\n\nAdd or update `LangBot/docs/event-based-agents/adapters/.md` in the same style as Telegram:\n\n- Status and adapter directory.\n- Configuration table matching manifest fields.\n- Supported EBA event list.\n- Common API table with support and limitations.\n- `call_platform_api` action list.\n- Receive component table with evidence level per component.\n- Send component table with evidence level per component.\n- Event table with evidence level per event.\n- Common API table with evidence level per API.\n- Platform-specific API table with evidence level per action.\n- Live test record with exact date, endpoint/simulator, standalone runtime command, test plugin path/name, JSONL evidence path, channel/group type, observed events, APIs exercised, destructive operations, and skipped items.\n\nBe honest. Put untested or skipped APIs in the document with the reason. Do not imply full parity when a platform cannot provide the same information density.\n\n## Before Finishing\n\n- Run unit tests and compile the live probe.\n- Run the standalone runtime plugin E2E path for every required acceptance item that the platform supports.\n- Run `git diff --check`.\n- Summarize live JSONL evidence by event type.\n- Stop all long-running runtimes and probes.\n- Confirm no secrets are staged.\n- Leave unrelated untracked files alone.\n","skills/skills/langbot-env-setup/SKILL.md":"---\nname: langbot-env-setup\ndescription: Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths.\n---\n\n# LangBot Environment Setup\n\nUse this skill when a task needs LangBot to be in a testable state before product testing or development verification.\n\n## Routing\n\n- **Shared local variables**: read `../.env` before using URL, path, browser profile, or proxy defaults.\n- **Always start here**: read `references/browser-access-selection.md` to choose the browser-control path.\n- **LangBot service checks and startup**: read `references/service-startup.md`.\n- **Computer Use available**: read `references/computer-use.md`. This path usually needs less browser/MCP setup.\n- **No Computer Use, browser automation required**: read `references/playwright-mcp.md`.\n- **GitHub OAuth or persisted login profile**: read `references/oauth-browser-profile.md`.\n- **WSL-specific notes**: read `references/wsl-notes.md` only when running under WSL.\n- **Proxy setup**: read `references/proxy.md` when external login, model provider tests, or package downloads time out.\n- **Headless-only automation**: use only after a profile already contains a valid LangBot login. Do not ask the agent to enter GitHub credentials or 2FA.\n\n## Rules\n\n- Never handle the user's GitHub password, passkey, recovery code, or 2FA secret.\n- For OAuth login, open a visible browser and let the user complete the credential steps.\n- Reuse a fixed browser profile path so the agent can later access the logged-in LangBot session.\n- Keep environment-specific paths and commands in `references/`, not in this file.\n- Treat environment setup as complete only after the target LangBot services are reachable and the browser profile can access the WebUI.\n","skills/skills/langbot-env-setup/references/browser-access-selection.md":"# Browser Access Selection\n\nChoose the lightest browser-control path that can complete the task.\n\n## Decision Order\n\n1. If Codex Computer Use, Claude Computer Use, or another visible browser-control tool is available, use `computer-use.md`.\n2. If no computer-control tool is available but Playwright MCP is available, use `playwright-mcp.md`.\n3. If the browser session must survive restarts or OAuth login is required, also use `oauth-browser-profile.md`.\n4. If running under WSL, add `wsl-notes.md`.\n5. If external sites or model providers time out, add `proxy.md`.\n\n## Principle\n\nComputer Use and Playwright MCP are alternative browser-control paths. Both still need LangBot services to be reachable, so service checks stay in `service-startup.md`.\n","skills/skills/langbot-env-setup/references/computer-use.md":"# Computer Use Browser Path\n\nUse this path when Codex Computer Use, Claude Computer Use, or another agent-visible browser-control capability is available.\n\n## Why This Path Is Simpler\n\nComputer Use can interact with a visible browser directly, so it usually does not need Playwright MCP configuration or a separate MCP browser bridge.\n\n## Workflow\n\n1. Verify LangBot backend/frontend with `service-startup.md`.\n2. Open the WebUI in the controlled browser.\n3. If login is needed, let the user complete GitHub OAuth. Never handle credentials or 2FA.\n4. Keep the browser/profile available for later testing.\n5. Hand off to `langbot-testing` after the page shows the logged-in WebUI.\n\n## Still Required\n\n- Proxy may still be needed for GitHub OAuth or model provider tests. Use `proxy.md`.\n- Persisted profile details may still matter if the computer-control browser is restarted. Use `oauth-browser-profile.md` if login state must survive.\n","skills/skills/langbot-env-setup/references/oauth-browser-profile.md":"# OAuth Browser Profile\n\nUse this reference when LangBot or LangBot Space needs GitHub OAuth login and the agent must reuse the authenticated browser state later.\n\nRead `skills/.env` first for `LANGBOT_BACKEND_URL`, `LANGBOT_FRONTEND_URL`, `LANGBOT_BROWSER_PROFILE`, `LANGBOT_CHROMIUM_EXECUTABLE`, and proxy defaults.\n\n## Rules\n\n- Never handle the user's GitHub password, passkey, recovery code, or 2FA secret.\n- Open a visible browser and let the user complete credential steps.\n- Reuse a fixed browser profile path.\n- Do not print token values. It is acceptable to report localStorage key names.\n\n## Manual Visible Login Flow\n\n1. Verify LangBot backend is reachable with `service-startup.md`.\n2. Launch a visible Chromium window with the persistent profile:\n\n```bash\nsetsid \"$LANGBOT_CHROMIUM_EXECUTABLE\" \\\n --no-sandbox \\\n --ozone-platform=x11 \\\n --user-data-dir=\"$LANGBOT_BROWSER_PROFILE\" \\\n --proxy-server=\"$LANGBOT_PROXY_SOCKS\" \\\n --proxy-bypass-list=\"$LANGBOT_NO_PROXY\" \\\n \"$LANGBOT_BACKEND_URL/login\" \\\n >/tmp/langbot-visible-chrome.log 2>&1 < /dev/null &\n```\n\n3. The user completes:\n\n```text\nLogin with Space -> Login with GitHub -> GitHub credentials / 2FA -> authorize\n```\n\n4. The agent can then reuse the same profile for automated checks.\n\n## Expected Successful State\n\nAfter login, LangBot should redirect away from `/login`, for example to a `/home/...` URL on the selected origin.\n\nExpected visible signals:\n\n```text\nLangBot\nDashboard\nHome\nBots\nPipelines\nKnowledge\nExtensions\n```\n\nExpected localStorage key names:\n\n```text\ntoken\nuserEmail\nlangbot_language\n```\n\nIf the user logged in on one origin but `LANGBOT_FRONTEND_URL` still shows `/login`, copy only the auth state needed between origins. Do not print token values.\n","skills/skills/langbot-env-setup/references/playwright-mcp.md":"# Playwright MCP Browser Path\n\nUse this path when the agent needs browser automation but no Computer Use browser-control path is available.\n\n## Known Paths\n\n- Persistent browser profile: `LANGBOT_BROWSER_PROFILE` from `skills/.env.local`\n- Chromium executable: `LANGBOT_CHROMIUM_EXECUTABLE` from `skills/.env.local`\n- Codex MCP config: `$CODEX_HOME/config.toml` or the config path used by the active agent.\n\n## MCP Config\n\nKeep the profile path fixed so the agent can reuse authenticated state.\n\n```toml\n[mcp_servers.playwright]\ncommand = \"npx\"\nargs = [\"-y\", \"@playwright/mcp@latest\", \"--no-sandbox\", \"--executable-path\", \"\", \"--proxy-server\", \"\", \"--proxy-bypass\", \"localhost,127.0.0.1\", \"--user-data-dir\", \"\"]\n```\n\nAfter changing MCP config, restart Codex so the MCP server is relaunched with the new args.\n\n## Visible Login\n\nFor OAuth login, Playwright MCP's headless browser is not enough. Launch a visible browser with the same profile and let the user complete login. Use `oauth-browser-profile.md`.\n\n## Common Failures\n\n- MCP still uses old args after editing config: restart Codex or kill old `playwright-mcp` processes and restart the session.\n- Browser is headless during OAuth: use the visible login command from `oauth-browser-profile.md`.\n","skills/skills/langbot-env-setup/references/proxy.md":"# Proxy Setup\n\nUse this reference when GitHub OAuth, package installation, model provider tests, or external API calls time out.\n\nRead defaults from `skills/.env` first.\n\n## Standard Local Proxy\n\n```bash\nexport HTTP_PROXY=\"$LANGBOT_PROXY_HTTP\"\nexport HTTPS_PROXY=\"$LANGBOT_PROXY_HTTP\"\nexport ALL_PROXY=\"$LANGBOT_PROXY_SOCKS\"\nexport http_proxy=\"$LANGBOT_PROXY_HTTP\"\nexport https_proxy=\"$LANGBOT_PROXY_HTTP\"\nexport all_proxy=\"$LANGBOT_PROXY_SOCKS\"\nexport NO_PROXY=\"$LANGBOT_NO_PROXY\"\nexport no_proxy=\"$LANGBOT_NO_PROXY\"\n```\n\n## Rule\n\nKeep uppercase and lowercase proxy variables consistent. Different libraries read different names.\n\n## Checks\n\n```bash\nenv | rg -i '^(http|https|all|no)_?proxy='\ncurl -I --max-time 8 --proxy \"$LANGBOT_PROXY_SOCKS\" https://github.com\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL\"\n```\n","skills/skills/langbot-env-setup/references/service-startup.md":"# Service Startup\n\nUse this reference for LangBot backend/frontend readiness checks regardless of OS or browser-control method. Read `skills/.env` first and override those defaults with user-provided values or detected running services.\n\n## Variables\n\n- `LANGBOT_REPO`\n- `LANGBOT_WEB_REPO`\n- `LANGBOT_BACKEND_URL`\n- `LANGBOT_FRONTEND_URL`\n- `LANGBOT_DEV_FRONTEND_URL`\n\n## Backend\n\nStart LangBot from the backend repo:\n\n```bash\ncd \"$LANGBOT_REPO\"\nuv run main.py\n```\n\nHealthy startup includes:\n\n```text\nRunning on http://0.0.0.0:\nConnected to plugin runtime.\nPlugin langbot-team/LocalAgent initialized\n```\n\nQuick check:\n\n```bash\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL/login\"\n```\n\nIf `bin/lbs env doctor` reports that `LANGBOT_BACKEND_URL` has no TCP listener,\nthe backend is not running at the configured host and port. A reachable\nstandalone frontend on `LANGBOT_FRONTEND_URL` does not prove backend readiness.\n\nPrefer a visible terminal session while debugging backend startup. Detached\nbackground startup methods can hide early process exits in local agent runs; if\nyou use one, immediately verify both the process and the listener:\n\n```bash\nps -eo pid,cmd | rg 'main.py|uv run main|langbot'\nss -ltnp | rg ':5300'\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL/login\"\n```\n\n## Frontend\n\nStart the new frontend from the web repo:\n\n```bash\ncd \"$LANGBOT_WEB_REPO\"\nVITE_API_BASE_URL=\"$LANGBOT_BACKEND_URL\" pnpm dev --host 0.0.0.0\n```\n\nHealthy startup includes:\n\n```text\nLocal: \n```\n\nQuick check:\n\n```bash\ncurl -I --max-time 3 \"$LANGBOT_FRONTEND_URL\"\n```\n\nIf `VITE_API_BASE_URL` is missing, Vite still serves the page but frontend API\ncalls may go to the frontend port instead of the backend port. That produces\nfalse browser failures in login, wizard, pipeline, and Debug Chat cases.\n\n## Completion Signal\n\nEnvironment setup is not complete until the required frontend/backend URLs are reachable and the chosen browser-control path can open the WebUI.\n","skills/skills/langbot-env-setup/references/wsl-notes.md":"# WSL Notes\n\nUse this reference only for WSL-specific details. Do not put generic LangBot startup or browser-login steps here.\n\n## Network\n\nGitHub login and model provider calls may require proxy access from WSL.\n\nWorking proxy form:\n\n```bash\nsocks5://127.0.0.1:7890\n```\n\nBypass local LangBot:\n\n```bash\nlocalhost,127.0.0.1\n```\n\nQuick checks:\n\n```bash\ncurl -I --max-time 8 --proxy socks5h://127.0.0.1:7890 https://github.com\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL\"\n```\n\n## Visible Browser\n\nIf OAuth requires a visible browser, WSL must have a usable display path. If a visible Chromium launch fails, check the local WSL GUI/X11 setup before changing LangBot config.\n\n## Common Failures\n\n- `ERR_NETWORK_CHANGED` or GitHub timeout: browser is not using the SOCKS proxy.\n- LangBot connection refused: backend is not running or not reachable from WSL.\n- User cannot type credentials: browser is headless or not visible to the user.\n","skills/skills/langbot-mcp-ops/SKILL.md":"---\nname: langbot-mcp-ops\ndescription: Operate a LangBot instance through its built-in MCP (Model Context Protocol) server. Use when an AI agent needs to manage LangBot — list/create/update/delete bots, agents, pipelines, models, knowledge bases, MCP servers, and skills — over MCP instead of raw HTTP. Covers the /mcp endpoint, API-key auth (web-UI lbk_ keys and the config.yaml global key), the tool surface, and client configuration. Triggers on \"langbot mcp\", \"manage langbot via mcp\", \"langbot /mcp\", \"langbot mcp server\".\n---\n\n# LangBot MCP Operations\n\nLangBot exposes an **MCP server** so AI agents can manage an instance\nprogrammatically. It mirrors a curated subset of the HTTP service API.\n\n## Endpoint\n\n```\nhttp://:5300/mcp\n```\n\nTransport: **streamable HTTP** (stateless, JSON responses). Same host/port as\nthe web UI and HTTP API.\n\n## Authentication\n\nReuses the same API keys as the HTTP API. Send either header:\n\n```\nX-API-Key: \n# or\nAuthorization: Bearer \n```\n\nTwo kinds of key are accepted:\n\n1. **Web-UI key** — created in the web UI (sidebar → API Keys), prefixed `lbk_`.\n The secret is shown once; only its SHA-256 hash is stored. Each key is bound\n to one Workspace and has explicit scopes, status, optional expiry, and\n last-used metadata. The key determines the Workspace; callers cannot switch\n it with `X-Workspace-Id`.\n2. **Global API key** — set in `data/config.yaml` under `api.global_api_key`.\n Requires no login session and no DB record; does not need the `lbk_` prefix.\n It is accepted only by a community instance with exactly one local\n Workspace and is disabled for SaaS multi-Workspace operation. Leave empty to\n disable. See the `langbot-deploy` skill for config details.\n\nInvalid, revoked, or expired keys get `401 Unauthorized`. A valid key whose\nscopes do not authorize a tool gets `403 Forbidden`.\n\nTo inspect key identity and permissions, call `GET /api/v1/system/context` with the API key.\n\n## Client configuration\n\n```json\n{\n \"mcpServers\": {\n \"langbot\": {\n \"url\": \"http://:5300/mcp\",\n \"headers\": { \"X-API-Key\": \"\" }\n }\n }\n}\n```\n\n## Tool surface\n\nSlack quick setup: create a disabled `slack-omni` bot draft first, then use\n`start_slack_setup` with the user's App Configuration access/refresh tokens.\nChoose `socket_mode=true` with a browser-reachable OAuth `redirect_url`, or\nprovide the public HTTPS `webhook_url` ending in `/bots/`.\nPoll `get_slack_setup_status` for the Slack installation authorization URL;\nthe user must authorize installation. On success, apply the returned `config`\nusing the normal bot update flow. Socket Mode also requires a separately\ngenerated `app_token` (`xapp-`, scope `connections:write`). Never log tokens.\n`cancel_slack_setup` removes the temporary session without deleting the Slack\napplication. Sessions expire after 15 minutes and are bound to the initiating\nWorkspace, principal, and placement generation. Restarting a failed setup can\ncreate another Slack application; inspect the returned `app_id` first.\n\nThe tools wrap the LangBot service layer. Current tools (v1):\n\n`create_pipeline` and HTTP `POST /api/v1/pipelines` preserve a supplied `config`\nand `extensions_preferences`; omit `config` to use installed Runner defaults.\nSelecting `ai.runner.id` never grants extension access: an empty `plugins` list\nwith `enable_all_plugins=false` still denies that plugin's Tool, EventListener\nand Command components, while allowing its selected Runner. Editing and copying\na pipeline retain these independent settings.\n\n| Tool | Purpose |\n| --- | --- |\n| `get_system_info` | Version, edition, instance id |\n| `list_bots` / `get_bot` / `create_bot` / `update_bot` / `delete_bot` | Manage messaging-platform bots (secrets redacted on read) |\n| `list_bot_event_route_statuses` | Inspect bot event-route runtime status |\n| `list_processors` / `get_processor` / `create_processor` / `update_processor` / `delete_processor` | Manage the peer Agent, Pipeline and Event processor types |\n| `get_processor_metadata` | Discover installed event-capable Runner components, schemas and supported event patterns. |\n| `list_processor_runs` / `get_processor_run_events` | Read one Agent or plugin processor run history and logs; paginate with `before_id` / `after_sequence`. |\n| `debug_agent` | Execute a synthetic Agent event (`processor_uuid`, `payload`); requires `runtime.operate`. Returns final text and up to 1000 execution events (thinking, text, tool arguments/results). Platform tools use Mock; other configured tools execute normally. Optional `payload.mock`: `errors`/`results` keyed by platform tool name, `unsupported_apis` lists unavailable platform APIs. |\n| `list_pipelines` / `get_pipeline` / `create_pipeline` / `update_pipeline` / `delete_pipeline` | Manage pipelines |\n| `list_llm_models` / `get_llm_model` / `list_embedding_models` / `list_model_providers` | Inspect models & providers |\n| `list_knowledge_bases` / `get_knowledge_base` / `retrieve_knowledge_base` | RAG knowledge bases (incl. semantic search) |\n| `list_mcp_servers` | External MCP servers LangBot connects to (as a client) |\n| `list_skills` / `get_skill` | Installed skills |\n| `list_knowledge_engines` / `get_knowledge_engine_schema` / `list_knowledge_parsers` | Discover RAG configuration |\n| `get_pipeline_extensions` / `update_pipeline_extensions` | Read or completely replace extension bindings; all lists and switches required |\n| `run_pipeline` | One fresh-session turn; requires `runtime.operate`, executes configured models/tools, never auto-retry an unknown outcome |\n| `get_monitoring_records` / `get_monitoring_details` | Bounded Workspace records and existing message/session details |\n| `get_sandbox_diagnostics` | Read status (`resource.view`), sessions/errors (`audit.view`); managed sandbox admission still applies |\n\nMutating tools (`create_*`, `update_*`) take a JSON object matching the same\nshape as the corresponding HTTP API request body. Discover resources with the\n`list_*` / `get_*` tools before mutating; identifiers are UUIDs. Reads require\n`resource.view`; mutations require `resource.manage`. All service calls inherit\nthe immutable Workspace context authenticated at the MCP transport boundary.\nPass `is_default: true` to `create_pipeline` only when the Workspace does not\nalready have a default pipeline.\n\n## How to use\n\nPipeline plugin extension switches authorize only EventListener, Tool and Command\ncomponents. Select the Runner independently in AI settings. An empty extension\nallowlist does not prevent saving, copying or running that Runner. Never enable\nall plugins or add its plugin to the allowlist as a Runner workaround: that also\nauthorizes its other components. The Runner plugin must still be installed and\nenabled in the authenticated Workspace, with an available Plugin Runtime.\n\n1. Get an API key (web UI key, or set `api.global_api_key` in config.yaml).\n2. Point your MCP client at `http://:5300/mcp` with the key header.\n3. Call `get_system_info` to confirm connectivity.\n4. Use `list_*` tools to discover, then `get_*` / `create_*` / `update_*` /\n `delete_*` as needed.\n\n## ChatGPT / Codex subscription providers\n\n`list_model_providers` can return the `openai-codex` requester. Its OAuth\ncredentials are server-only and are not provider API keys. Never ask a user\nto paste ChatGPT access tokens, refresh tokens, or a Codex auth cache into an\nMCP tool or model configuration.\n\nA human connects or disconnects the subscription through **Models → provider\nsettings** in the LangBot web UI. The provider-scoped `/codex/*` authentication\nroutes deliberately require a browser-user session and are not exposed as MCP\ntools or authorized by a LangBot API key. Once connected, models are managed\nand selected through the normal provider/model workflow. A disconnected\nprovider must be reauthorized; do not silently replace it with API-key billing.\n\nSee [ChatGPT / Codex subscription](../../../docs/CODEX_SUBSCRIPTION.md) for setup,\nusage limits, and the personal-account versus shared-service boundary.\n\n## Provider deletion\n\nThe curated MCP surface currently lists providers but has no provider-deletion\ntool. In the web UI, **Edit Provider → Delete** asks for confirmation before\nremoving that provider and all its LLM, embedding, and rerank models. This is\nirreversible; never interpret a request to edit a provider as authorization to\ndelete it.\n\nThe equivalent HTTP operation is\n`DELETE /api/v1/provider/providers/{uuid}?cascade=true`, requiring\n`resource.manage` in the authenticated Workspace. Omitting `cascade` preserves\nthe existing refusal to delete providers that still have models. Cloud-managed\nproviders remain protected. Cascade deletion removes stored Codex authorization\nstate as well; it is not the same operation as disconnecting an account.\n\n## Implementation & maintenance (for LangBot developers)\n\n- Server: `src/langbot/pkg/api/mcp/server.py` (FastMCP). Tools call the service\n layer directly, so the MCP surface stays aligned with the API.\n- Mount: `src/langbot/pkg/api/mcp/mount.py` — an ASGI dispatcher fronting Quart,\n authenticating `/mcp` requests, running the streamable-HTTP session manager.\n- Smoke test: `tests/manual/mcp_smoke.py`.\n\n> When you add, remove, or change an HTTP API endpoint that should be\n> agent-accessible, update the corresponding MCP tool **and** this skill. The\n> MCP tool surface and the API must stay aligned (see `AGENTS.md`).\n\n## Pitfalls\n\n- `/mcp` is the **server** LangBot exposes. The `/api/v1/mcp` routes are the\n **client** side (managing external MCP servers LangBot connects to). Don't\n confuse them.\n- A `401` means the key is wrong, missing, revoked, expired, or (for the global\n key) `api.global_api_key` is empty or the instance is not an OSS singleton.\n- A `403` means the key is valid but lacks the permission required by the tool.\n- The global key is plaintext in config.yaml — only enable it on trusted/internal\n deployments and serve over HTTPS.\n\n## Default message routes\n\n`create_bot` without non-empty `event_bindings` adds two explicit message routes:\nprivate chat (`chat_type eq person`) and group chat (`chat_type eq group` plus\n`is_at_bot eq true`). Both use the Workspace's default Pipeline, or the first\navailable Pipeline. If none exists, first-bot setup creates one with the full\ndefault config. Creating a Pipeline via `create_agent(kind=\"pipeline\")` also uses\nthat full default config. Configure an authorized model before testing replies.\n\nExplicit routes on create are preserved; `update_bot` never automatically\nreplaces manual routes, including an intentionally empty list. The WebUI message\ntemplate uses the same private/group conditions and preserves existing routes\nwhen applied repeatedly. Unmentioned group messages do not enter the Pipeline.\nAn existing broad route can still match first; inspect route order and diagnostics\nbefore changing a member's configuration.\n\n## Event processors\n\nCreate a processor with `kind: \"event_processor\"` and basic information. Without\na component it supports no events. Discover installed components with\n`get_processor_metadata`, then use `update_processor` with `component_ref` and\noptional `parameters`. API callers may also supply these when creating an instance.\nBind an instance by updating the bot's `plugin_processors` array with\n`{\"processor_uuid\": \"\", \"enabled\": true}`. This replaces the full\nsubscription list; preserve bindings you want to keep. Do not add plugin processors\nto `event_bindings`, which remains exclusive Agent/Pipeline routing.\nEach enabled subscription independently receives the installed Runner's declared\nevents. Slow or failed subscribers do not prevent other subscribers or the primary\nroute from executing. Installation alone never activates a handler. Reusing an\ninstance shares its configuration and runtime state. Use a separate instance for\nindependent settings. Optional plugin behavior belongs in the Runner config schema.\n`debug_agent` accepts the complete typed event in `payload.data` for this kind.\nLegacy EventListener plugins remain in the Pipeline lifecycle.\n\n`list_processor_runs` includes `created_at_ms`, `started_at_ms`, and\n`finished_at_ms`: Host lifecycle times in epoch milliseconds. Use the start and finish times for elapsed processing time; select a run and call `get_processor_run_events` for its\nlogs and action results. These times are not internal plugin profiling data.\n\n## Unified execution monitoring\n\nUse `get_inflight_executions` for a bounded current-workspace snapshot of active\nand recently started executions, including `progress_event`. Progress is a\nreported stage, not an estimated completion percentage. The UI uses the shared\n`GET /api/v1/monitoring/in-flight/stream` SSE feed instead of polling per viewer.\n\nUse `get_monitoring_executions` for the execution list and\nits legacy `pipeline_ids` parameter to filter any processor kind (Agent,\nPipeline or event processor); the summary uses the same processor scope. Use\n`get_monitoring_execution_detail` with `source=auto` to resolve a run,\nmessage or event identifier outside the current list page. The detail exposes\n`inputs`, `outputs` (generated content), `deliveries` (recorded platform sends),\n`events`, `llm_calls`, `tool_calls`, `errors`, `related`, and `conversation` in\n`pages`. Follow each section's `has_more` and `next_offset` independently.\nStart with `section=overview` to load only inputs, outputs, deliveries and errors.\nRequest `conversation`, `related`, calls and events separately when needed;\nomitting `section` retains the complete trace response for existing consumers.\nConversation history supplies context; historical messages without explicit\nlinks must not be asserted to belong to the selected execution. Events without\na processor run use `source=event`. All lookups remain Workspace-scoped.\n\nMonitoring record filters, except for `kind=\"sessions\"`, accept `mode` (`all`, `real`, `debug`) and\n`execution_statuses` (normalized execution statuses). These select the owning\nexecution, not the individual model/tool call outcome. Calls without a recorded\nexecution link are excluded when an execution filter is active.\n\nSession lists support `bot_ids`, `pipeline_ids`, `start_time`, `end_time`,\n`user_query`, `is_active`, and pagination. For execution status or real/debug\nfilters, use `get_monitoring_executions`; HTTP `/monitoring/sessions` rejects\nthem. In HTTP `/monitoring/data`, `sessions` and `totalCount.sessions` use only\nbot/pipeline/time filters; execution filters apply to the other records.\n\n### Workspace message language\n\n`get_workspace_message_language` reads the current Workspace's backend language\n(`workspace.view`). `set_workspace_message_language(language)` requires\n`workspace.update`; supported values are `en_US`, `zh_Hans`, `zh_Hant`, `ja_JP`,\n`es_ES`, `ru_RU`, `th_TH`, `vi_VN`. HTTP equivalents are GET/PUT\n`/api/v1/workspaces/current/language` (PUT body: `{\"language\":\"ja_JP\"}`).\nThe API key's Workspace remains authoritative. The setting controls new backend\nmessages and default bot notices, independently of personal UI language.\nPipeline `output.misc` hint fields override defaults; empty values inherit this\nsetting. Existing custom text and historical logs are not rewritten.\n\n### Workspace default LLM model\n\n`get_starred_model` reads the Workspace-wide favorite. `set_starred_model`\nrequires `provider_secret.manage` and replaces the single starred model; pass\n`null` to clear it. The model must belong to the current Workspace.\n`get_default_model` resolves the favorite first, then the Space wizard chat\nrecommendation only when LangBot Models is enabled and the Workspace owner is\nbound to a LangBot Account. It returns a nullable `uuid`.\nNew processor LLM selector defaults use this preference; existing configurations\nare not rewritten. Embedding and rerank models are not eligible.\n\n### Reset a bot session context\n\nUse `reset_session_context(bot_id, session_id)` only when a user asks to start a session afresh. It requires `resource.manage`, preserves monitoring records, and refuses sessions with active tasks. It clears conversation runner state and excludes earlier transcript entries from subsequent model context. Files and long-term memory are not removed.\n\n## Cloud plan platform restrictions\n\nBot creation and enabling use the same Workspace entitlement policy as HTTP. A disabled platform is rejected even through MCP or plugin calls; do not retry using an Omni or legacy alias to bypass the restriction. The Workspace-scoped HTTP adapter catalog omits disabled platforms.\n\n## LLM request diagnostics\n\n`get_monitoring_records(kind=\"llm_calls\")` includes nullable `details` on each call.\nRead `details.timing`, `details.usage`, and `details.error` to distinguish provider\nlatency, retry waits, cached/reasoning tokens, and failures. Timing values are\nmilliseconds since invocation start; `first_event_ms` includes reasoning/control\nframes, while `first_content_ms` measures the first upstream content delta.\n`first_visible_chunk_ms` may contain displayed reasoning, not just answer text.\nOld records have no details. Do not equate total duration with provider TTFT.\nSee `docs/LLM_REQUEST_POLICY.md` for policy precedence and gateway header behavior.\n\n`get_llm_model` also exposes optional `reasoning_capabilities.budget_tokens`\nmin/max bounds. Numeric thinking budgets use `reasoning_config.budget_tokens`\nwith level `enabled`; do not combine budgets with effort levels or native thinking\narguments. Existing Runner reasoning overrides take precedence. Prompt caching\nis opt-in through model/call `extra_args.prompt_caching`; use provider-specific\nmodes documented in `docs/LLM_REQUEST_POLICY.md`. Cache keys are routing hints,\nnot evidence of hits; read the actual usage counters. `preparation_ms` measures\nrequest construction and `preparation_cache_hits` counts retry preparation reuse.\n`transport_setup_ms` measures managed pool acquisition, including first-use TLS\ninitialization; it is not network TTFT. See `docs/LLM_LATENCY_BENCHMARK.md` for\ncontrolled request-layer measurements and their scope.\n","skills/skills/langbot-plugin-dev/SKILL.md":"---\nname: langbot-plugin-dev\ndescription: Develop, debug, and test LangBot plugins. Use when creating new LangBot plugins, fixing plugin bugs, setting up a LangBot test environment, or testing plugins via WebSocket. Covers plugin component architecture (EventListener, Command, Tool), the plugin SDK API (invoke_llm, get_llm_models, send_message, plugin storage), common pitfalls, and automated WebSocket-based testing. Triggers on \"langbot plugin\", \"lbp\", \"GroupChatSummary\", \"plugin debug\", \"langbot test\".\n---\n\n# LangBot Plugin Development & Debugging\n\n## Controlling a running instance via MCP\n\nBeyond writing code, you can **drive a live LangBot instance over MCP** — no raw\nHTTP needed. Two MCP servers exist (both reuse existing API keys; see `AGENTS.md`):\n\n- **LangBot instance** — `http://:5300/mcp` (auth: web-UI `lbk_` key or the\n `api.global_api_key` from `config.yaml`). Manage bots, pipelines, models,\n knowledge bases, and skills. See the **`langbot-mcp-ops`** skill.\n- **LangBot Space marketplace** — `https://space.langbot.app/mcp` (auth: Personal\n Access Token). Search plugins / MCP servers / skills. See the\n **`langbot-space-ops`** skill.\n\n> Any change to an agent-accessible HTTP API endpoint must keep the matching MCP\n> tool and these skills in sync.\n\n## Plugin Architecture\n\nA LangBot plugin consists of:\n\n```\nMyPlugin/\n├── manifest.yaml # Plugin metadata, config schema\n├── main.py # BasePlugin subclass (entry point, shared state)\n├── components/\n│ ├── event_listener/ # Hook pipeline events\n│ │ ├── collector.yaml\n│ │ └── collector.py\n│ ├── commands/ # !command handlers\n│ │ ├── mycommand.yaml\n│ │ └── mycommand.py\n│ └── tools/ # LLM function-call tools\n│ ├── mytool.yaml\n│ └── mytool.py\n```\n\nEach component has a `.yaml` (metadata) and `.py` (implementation).\n\n## README & i18n convention (enforced on the marketplace)\n\nA plugin published to LangBot Space serves a localized README on its detail page.\nThe resolver (`langbot-space` `PluginService.GetPluginREADME`) works like this:\n\n- **Root `README.md` MUST be in English.** It is the default and the fallback —\n when no per-language README matches the viewer's locale, the page serves the\n root `README.md`. A non-English root README makes the English/default view show\n the wrong language.\n- **All other languages live under `readme/README_{lang}.md`** — e.g.\n `readme/README_zh_Hans.md`, `readme/README_ja_JP.md`. The 8 supported locales:\n `en_US, zh_Hans, zh_Hant, ja_JP, th_TH, vi_VN, es_ES, ru_RU`.\n- `manifest.yaml` `metadata.label` / `metadata.description` should carry the same\n 8-locale i18n set (`repository` must be a real, alive URL).\n\n```\nMyPlugin/\n├── manifest.yaml\n├── README.md # English (default + fallback) — REQUIRED, must be English\n└── readme/\n ├── README_zh_Hans.md\n ├── README_zh_Hant.md\n ├── README_ja_JP.md\n ├── README_th_TH.md\n ├── README_vi_VN.md\n ├── README_es_ES.md\n └── README_ru_RU.md\n```\n\n`manifest.yaml` (incl. `repository`) is the source of truth — the marketplace\nsyncs from it, so edit the package and re-publish rather than patching live data.\n\n## Critical SDK Pitfalls\n\n### 1. MessageChain is a RootModel — iterate directly\n\n```python\n# ❌ WRONG — MessageChain has no .components attribute\nfor component in event.message_chain.components:\n\n# ✅ CORRECT — MessageChain is a Pydantic RootModel, iterate directly\nfor component in event.message_chain:\n```\n\n### 2. Message.content must be `list[ContentElement]` or `str`, not a single ContentElement\n\n```python\nfrom langbot_plugin.api.entities.builtin.provider import message as provider_message\n\n# ❌ WRONG — single ContentElement\nMessage(role=\"user\", content=ContentElement.from_text(\"hello\"))\n\n# ✅ CORRECT — list of ContentElement\nMessage(role=\"user\", content=[ContentElement.from_text(\"hello\")])\n\n# ✅ ALSO CORRECT — plain string\nMessage(role=\"user\", content=\"hello\")\n```\n\n### 3. invoke_llm does NOT accept timeout\n\n```python\n# ❌ WRONG\nawait self.invoke_llm(llm_model_uuid=uuid, messages=msgs, timeout=60)\n\n# ✅ CORRECT\nawait self.invoke_llm(llm_model_uuid=uuid, messages=msgs)\n```\n\n### 4. invoke_llm response.content can be str OR list\n\n```python\nresponse = await self.invoke_llm(...)\nif response.content:\n if isinstance(response.content, str):\n return response.content\n elif isinstance(response.content, list):\n parts = [e.text for e in response.content if hasattr(e, \"text\") and e.text]\n return \"\\n\".join(parts)\n```\n\n### 5. get_llm_models() returns UUIDs\n\n```python\n# Returns list[str] of model UUIDs\nmodels = await self.get_llm_models()\nmodel_uuid = models[0] # First available model UUID\n```\n\n**Known bug (v4.9.3):** The host handler may return `list[dict]` instead of `list[str]`. If you hit `TypeError: unhashable type: 'dict'` in `invoke_llm`, the fix is in `LangBot/src/langbot/pkg/plugin/handler.py` — change `'llm_models': llm_models` to `'llm_models': [m['uuid'] for m in llm_models]`.\n\n### 6. invoke_llm parameter is `llm_model_uuid`, NOT `model_uuid`\n\n```python\n# ❌ WRONG — will throw \"got an unexpected keyword argument\"\nawait self.invoke_llm(messages=msgs, model_uuid=uuid)\n\n# ✅ CORRECT\nawait self.invoke_llm(messages=msgs, llm_model_uuid=uuid)\n```\n\n### 7. prevent_default() alone does NOT block LLM response\n\nTo fully prevent the default LLM pipeline from responding when your EventListener handles the message, you must call **both**:\n\n```python\nevent_context.prevent_default() # Block default behavior\nevent_context.prevent_postorder() # Block subsequent plugins/pipeline\n```\n\nUsing only `prevent_default()` still allows the LLM to generate a response.\n\n### 8. get_plugin_storage / set_plugin_storage may throw KeyError: 'owner'\n\nThis is a version mismatch between the SDK and host. Wrap storage calls in try/except:\n\n```python\ntry:\n data = await self.get_plugin_storage(\"my_key\")\nexcept Exception:\n data = None # Fallback gracefully\n```\n\n### 9. Component YAML must have full structure, not just name/description\n\n```yaml\n# ❌ WRONG — will silently fail to register the component\nname: translator\ndescription:\n en_US: 'Does stuff'\n\n# ✅ CORRECT — full component YAML\napiVersion: v1\nkind: EventListener\nmetadata:\n name: translator\n label:\n en_US: Translator\nspec:\nexecution:\n python:\n path: translator.py\n attr: Translator\n```\n\n### 10. BasePlugin import path\n\n```python\n# ❌ WRONG\nfrom langbot_plugin.api.definition.base_plugin import BasePlugin\n\n# ✅ CORRECT\nfrom langbot_plugin.api.definition.plugin import BasePlugin\n```\n\n## Pipeline Events\n\nEvents the EventListener can hook (from most general to most specific):\n\n| Event | When |\n|---|---|\n| `GroupMessageReceived` | **Any** group message arrives (before trigger rules) |\n| `PersonMessageReceived` | **Any** private message arrives |\n| `GroupNormalMessageReceived` | Group message passes trigger rules, going to LLM |\n| `PersonNormalMessageReceived` | Private message going to LLM |\n| `GroupCommandSent` | Group message matched as command |\n| `PersonCommandSent` | Private message matched as command |\n| `NormalMessageResponded` | LLM generated a response |\n| `PromptPreProcessing` | About to build LLM context |\n\n**Key insight:** `*MessageReceived` fires for ALL messages regardless of trigger rules. `*NormalMessageReceived` only fires for messages that match the pipeline's trigger rules (e.g., @bot, prefix, random%). Use `*MessageReceived` for message collection/logging.\n\n## EventContext API\n\n```python\n@self.handler(events.GroupMessageReceived)\nasync def on_msg(event_context: context.EventContext):\n event = event_context.event\n event.launcher_id # Group ID\n event.sender_id # Sender ID\n event.message_chain # MessageChain (iterate directly)\n\n # Reply to the current conversation\n await event_context.reply(MessageChain([Plain(text=\"hello\")]))\n\n # Block default pipeline behavior\n event_context.prevent_default()\n\n # Block subsequent plugins\n event_context.prevent_postorder()\n```\n\n## Setting Up a Test Environment\n\n### Deploy via Docker (GitOps + Portainer)\n\nSee `references/test-env-setup.md` for full deployment steps.\n\nQuick summary:\n1. Create `docker-compose.yaml` in `server-deploy` repo\n2. Deploy via Portainer git repository method\n3. Set up admin account via `/api/v1/user/init` POST\n4. Configure LLM provider and model via API\n5. Copy plugin to `data/plugins/` directory\n\n### WebSocket Testing\n\nLangBot's WebUI chat uses WebSocket. Connect to test message flow:\n\n```\nws://:/api/v1/pipelines//ws/connect?session_type=group\n```\n\n- `session_type=group` for group chat simulation\n- `session_type=person` for private chat (always triggers pipeline)\n\n**Requires Origin header** to pass CORS:\n```javascript\nconst ws = new WebSocket(url, {\n headers: { Origin: 'https://your-langbot-domain' }\n});\n```\n\nSend messages:\n```json\n{\"type\": \"message\", \"message\": [{\"type\": \"Plain\", \"text\": \"hello\"}]}\n```\n\nReceive:\n- `{\"type\": \"connected\", ...}` — connection established\n- `{\"type\": \"user_message\", \"data\": {...}}` — echo of sent message\n- `{\"type\": \"response\", \"data\": {\"content\": \"...\", \"is_final\": true/false}}` — bot reply (streamed)\n\n### Group Trigger Rules\n\nGroup messages only enter the pipeline if trigger rules are met:\n\n```json\n{\n \"group-respond-rules\": {\n \"at\": true, // Respond when @bot\n \"prefix\": [\"ai\"], // Respond to messages starting with \"ai\"\n \"random\": 0.0, // Probability of responding to any message (0.0-1.0)\n \"regexp\": [] // Regex patterns\n }\n}\n```\n\nFor testing, set `random: 1.0` via PUT `/api/v1/pipelines/` to respond to all messages.\n\n**Important:** EventListener hooks like `GroupMessageReceived` fire regardless of trigger rules. Only the LLM processing (`GroupNormalMessageReceived` and beyond) requires trigger rules.\n\n### Plugin Hot-Reload\n\nThere is **no hot-reload**. After changing plugin files:\n\n```bash\ndocker restart \n# Wait ~5 seconds for plugin to re-mount\n```\n\nThe main LangBot container does NOT need restart for plugin changes — only the runtime container.\n\n## API Quick Reference\n\n### Admin Setup\n\n```bash\n# Initialize admin account (first time only)\ncurl -X POST $BASE/api/v1/user/init \\\n -H \"Content-Type: application/json\" \\\n -d '{\"user\":\"admin@test.com\",\"password\":\"test123\"}'\n\n# Login\ncurl -X POST $BASE/api/v1/user/auth \\\n -H \"Content-Type: application/json\" \\\n -d '{\"user\":\"admin@test.com\",\"password\":\"test123\"}'\n# Returns: {\"data\":{\"token\":\"eyJ...\"}}\n```\n\n### Provider & Model Setup\n\n```bash\n# Create provider\ncurl -X POST $BASE/api/v1/provider/providers \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d '{\"name\":\"MyProvider\",\"requester\":\"new-api-chat-completions\",\"base_url\":\"https://api.example.com/v1\",\"api_keys\":[\"sk-xxx\"]}'\n\n# Create LLM model\ncurl -X POST $BASE/api/v1/provider/models/llm \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d '{\"name\":\"gpt-4o-mini\",\"provider_uuid\":\"\",\"abilities\":[\"chat\",\"tool-use\"]}'\n\n# List models\ncurl $BASE/api/v1/provider/models/llm -H \"Authorization: Bearer $TOKEN\"\n```\n\n### Pipeline Config\n\n```bash\n# Get pipeline\ncurl $BASE/api/v1/pipelines -H \"Authorization: Bearer $TOKEN\"\n\n# Update pipeline (e.g., set model, modify trigger rules)\ncurl -X PUT $BASE/api/v1/pipelines/ \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d ''\n```\n\n## Plugin Config Types\n\nSupported `type` values in `manifest.yaml` `spec.config`:\n\n| Type | Description | Value |\n|---|---|---|\n| `string` | Text input | string |\n| `int` / `integer` | Number input | int |\n| `float` | Decimal input | float |\n| `bool` / `boolean` | Toggle | bool |\n| `select` | Dropdown (needs `options`) | string |\n| `prompt-editor` | Multi-line prompt editor | string |\n| `llm-model-selector` | LLM model picker UI | UUID string |\n| `bot-selector` | Bot picker UI | UUID string |\n\nExample — let users choose which model the plugin uses:\n\n```yaml\nspec:\n config:\n - name: model\n type: llm-model-selector\n label:\n en_US: 'LLM Model'\n zh_Hans: 'LLM 模型'\n description:\n en_US: 'Select the LLM model. Falls back to first available if not set.'\n zh_Hans: '选择 LLM 模型。未设置时使用第一个可用模型。'\n required: false\n```\n\nRead config in plugin code:\n\n```python\nmodel_uuid = self.get_config().get(\"model\")\n```\n\n## Container Restart Timing\n\nAfter plugin file changes, **only the runtime container needs restart**:\n\n```bash\ndocker restart langbot-test-runtime\n# Wait ~15 seconds before testing\n```\n\n**When to restart both (runtime first, then host):**\n- Added/removed Command or Tool components (host caches component lists)\n- Changed `manifest.yaml` structure\n\n```bash\ndocker restart langbot-test-runtime\nsleep 8\ndocker restart langbot-test\nsleep 8\n```\n\n**⚠️ Do NOT restart both simultaneously** — the host may connect before plugins are mounted, causing 502 errors or missing plugin registrations.\n\n## Debugging Checklist\n\nWhen a plugin doesn't work:\n\n1. **Check runtime logs**: `docker logs ` — look for mount/init errors\n2. **Check host logs**: `docker logs ` — look for pipeline processing errors\n3. **Verify plugin loaded**: `GET /api/v1/plugins` — should list your plugin\n4. **Test person mode first**: `session_type=person` always triggers pipeline, isolating trigger rule issues\n5. **Check trigger rules**: Group mode requires @bot, prefix match, or random% to enter pipeline\n6. **Verify model configured**: Pipeline's `config.ai.runner_config[config.ai.runner.id].model.primary` must point to a valid model UUID with working API keys\n\n## Publishing Plugins\n\nAfter testing, publish via `lbp publish`:\n\n```bash\ncd /path/to/MyPlugin\nlbp publish\n```\n\nThis builds `.lbpkg` and uploads to Space marketplace as a draft. Then go to https://space.langbot.app/market to upload screenshots and submit for review.\n\n**Prerequisite:** Must be logged in via `lbp login --token lbpat_xxx` (PAT from Space profile page).\n\n## Reference: EventListener-Only Plugin Pattern\n\nFor plugins that react to messages without commands or tools (e.g., auto-summarize URLs, collect messages, translate):\n\n```\nMyPlugin/\n├── manifest.yaml # Only EventListener in spec.components\n├── main.py # BasePlugin with shared logic (fetch, LLM calls)\n├── components/\n│ └── event_listener/\n│ ├── detector.yaml\n│ └── detector.py\n└── requirements.txt\n```\n\n**manifest.yaml** — only declare EventListener:\n```yaml\nspec:\n components:\n EventListener:\n fromDirs:\n - path: components/event_listener/\n```\n\n**detector.py** — hook `*MessageReceived`, extract text, process, reply:\n```python\n@self.handler(events.PersonMessageReceived)\nasync def on_msg(event_context: context.EventContext):\n event = event_context.event\n text_parts = []\n for component in event.message_chain:\n if isinstance(component, platform_message.Plain):\n text_parts.append(component.text)\n text = \"\".join(text_parts).strip()\n \n if should_handle(text):\n event_context.prevent_default()\n event_context.prevent_postorder()\n result = await self.plugin.process(text)\n await event_context.reply(platform_message.MessageChain([\n platform_message.Plain(text=result)\n ]))\n```\n\n**Key:** Access shared plugin logic via `self.plugin` (the BasePlugin instance).\n","skills/skills/langbot-plugin-dev/references/test-env-setup.md":"# Test Environment Setup\n\n## Docker Compose (GitOps)\n\nCreate in `server-deploy` repo under `servers//langbot-test/docker-compose.yaml`:\n\n```yaml\nversion: \"3\"\nservices:\n langbot_plugin_runtime:\n image: rockchin/langbot:latest\n container_name: langbot-test-runtime\n volumes:\n - /opt/docker-data/langbot-test/data/plugins:/app/data/plugins\n ports:\n - \"5411:5401\"\n restart: on-failure\n environment:\n - TZ=Asia/Shanghai\n command: [\"uv\", \"run\", \"--no-sync\", \"-m\", \"langbot_plugin.cli.__init__\", \"rt\"]\n networks:\n - langbot_test_network\n\n langbot:\n image: rockchin/langbot:latest\n container_name: langbot-test\n volumes:\n - /opt/docker-data/langbot-test/data:/app/data\n ports:\n - \"5310:5300\"\n restart: on-failure\n depends_on:\n - langbot_plugin_runtime\n environment:\n - TZ=Asia/Shanghai\n networks:\n - langbot_test_network\n\nnetworks:\n langbot_test_network:\n driver: bridge\n```\n\n## Post-Deploy Configuration\n\nAfter first start, LangBot auto-generates `data/config.yaml`. You need to update `plugin.runtime_ws_url` to match the runtime container name:\n\n```bash\n# On the host, edit config\nsed -i 's|ws://localhost:5400/control/ws|ws://langbot-test-runtime:5400/control/ws|' \\\n /opt/docker-data/langbot-test/data/config.yaml\ndocker restart langbot-test\n```\n\n## Installing a Plugin\n\nCopy plugin directory to `data/plugins/` on the host:\n\n```bash\nscp -r MyPlugin/ user@host:/opt/docker-data/langbot-test/data/plugins/MyPlugin/\ndocker restart langbot-test-runtime # Runtime picks up new plugins on restart\n```\n\n## Caddy Reverse Proxy (Optional)\n\nIf testing externally, add to Caddyfile on the same host:\n\n```\nlangbot-test.example.com {\n reverse_proxy langbot-test:5300\n}\n```\n\nThen reload: `docker exec caddy caddy reload --config /etc/caddy/Caddyfile`\n\nThe WebSocket endpoint works through Caddy without special config.\n\n## WebSocket Test Script (Node.js)\n\n```javascript\nconst WebSocket = require('ws');\n\nconst PIPELINE_UUID = '';\nconst BASE = 'wss://langbot-test.example.com';\nconst URL = `${BASE}/api/v1/pipelines/${PIPELINE_UUID}/ws/connect?session_type=group`;\n\nconst ws = new WebSocket(URL, {\n headers: { Origin: BASE }\n});\n\nconst send = (text) => {\n ws.send(JSON.stringify({\n type: 'message',\n message: [{ type: 'Plain', text }]\n }));\n console.log('[SENT]', text);\n};\n\nws.on('message', (data) => {\n const msg = JSON.parse(data.toString());\n if (msg.type === 'connected') {\n console.log('Connected!');\n // Send test messages\n send('Message 1');\n setTimeout(() => send('Message 2'), 500);\n setTimeout(() => send('!summary'), 2000);\n } else if (msg.type === 'response' && msg.data?.is_final) {\n console.log('[BOT]', msg.data.content);\n }\n});\n\nws.on('error', (e) => console.error('Error:', e.message));\nsetTimeout(() => { ws.close(); process.exit(); }, 60000);\n```\n\nRequires: `npm install ws`\n","skills/skills/langbot-skills-maintenance/SKILL.md":"---\nname: langbot-skills-maintenance\ndescription: Maintain the langbot-skills repository with low duplication. Use when adding, editing, or auditing LangBot skills, references, cases, troubleshooting entries, indexes, or periodic entropy-control checks for this skills repository.\n---\n\n# LangBot Skills Maintenance\n\nUse this skill before changing reusable assets in this repository.\n\n## Workflow\n\n1. Read `AGENTS.md`, `skills/.env`, and the relevant existing skill files.\n2. Classify the change:\n - `SKILL.md` for routing and concise operating rules.\n - `references/*.md` for canonical detailed workflows.\n - `cases/*.yaml` for executable test-plan skeletons.\n - `suites/*.yaml` for reusable groups of case ids.\n - `fixtures/fixtures.json` for deterministic fixture readiness metadata.\n - `reports/evidence//automation-result.json` as automation output and `reports/evidence//result.json` as final judgment output; neither is a catalog asset to commit.\n - `troubleshooting/*.yaml` for one reusable failure mode.\n3. Search existing assets before adding new files:\n - `rg \"\" skills`\n - `bin/lbs case list`\n - `bin/lbs suite list`\n - `bin/lbs fixture list`\n4. Put detail in one canonical place and link to it from cases or routing bullets.\n5. Run the checks in `AGENTS.md` after edits.\n\n## Entropy Rules\n\n- Prefer extending an existing reference or troubleshooting entry when the root cause is the same.\n- Keep cases short: setup, action, evidence, pass/fail checks. Do not paste long prompts or debug transcripts when a reference exists.\n- Put machine-checkable inputs in `env`, `automation_env`, or fixtures; put operator-confirmed assumptions in `preconditions` so `test plan` can surface `manual_check`.\n- Keep suites short: title, intent, tags, and ordered case ids. Do not duplicate case steps inside a suite.\n- Keep fixture manifests factual: id, title, path, kind, and related case ids. Do not encode environment-specific absolute paths.\n- Keep troubleshooting entries narrow: symptoms, patterns, likely causes, fixes, related assets.\n- Do not hardcode local ports, browser profile paths, secrets, tokens, or provider keys.\n- Use `bin/lbs index --check` to verify the committed index is current without writing it; run `bin/lbs index` when the index needs regeneration.\n\nFor periodic repository audits, read `references/curation-workflow.md`.\n","skills/skills/langbot-skills-maintenance/references/curation-workflow.md":"# Curation Workflow\n\nUse this checklist when the repository starts accumulating repeated cases, copied steps, or overlapping troubleshooting entries.\n\n## Audit Pass\n\n1. Inspect the current surface:\n - `bin/lbs case list`\n - `bin/lbs case list --json --priority p0 --automation`\n - `bin/lbs case list --ready`\n - `bin/lbs case list --machine-ready`\n - `bin/lbs suite list`\n - `bin/lbs fixture list`\n - `rg \"sandbox|provider|pipeline|plugin|knowledge|mcp\" skills`\n - `rg \"If .* fails|Known Pitfalls|Debug Chat|/api/v1\" skills`\n2. Group nearby assets by intent, not by file path:\n - user-facing scenario\n - backend or provider dependency\n - failure signature\n - pass/fail evidence\n3. Pick one canonical owner:\n - stable procedures belong in `references/`\n - deterministic files and packages belong in `fixtures/` plus `fixtures/fixtures.json`\n - repeated failure signatures belong in `troubleshooting/`\n - runnable QA paths belong in `cases/`\n - reusable groups of QA paths belong in `suites/`\n - skill entry points belong in `SKILL.md`\n\n## Merge Or Split\n\nMerge when two files share the same trigger, root cause, and fix. Keep the stronger id and move missing patterns into it.\n\nSplit when a file mixes unrelated failure modes or requires different fixes. Each troubleshooting id should map to one diagnosis path.\n\nMove repeated step lists out of cases and into a reference when more than one case would need the same prompt, UI path, or log interpretation.\n\nAdd or update a suite when developers repeatedly run the same ordered group of cases. Do not copy case steps into suites; use `bin/lbs suite plan ` to expand the group.\nUse `bin/lbs suite start ` and `bin/lbs suite report --evidence-dir ` when validating that a suite is operational end to end.\n\nAdd or update `fixtures/fixtures.json` when a case depends on a deterministic file, plugin package, or local test server. The manifest should use repo-relative paths under the owning skill and should not contain machine-local absolute paths.\n\nWhen adding Debug Chat Playwright automation, reuse `scripts/e2e/lib/debug-chat.mjs` for navigation, prompt send, response leaf matching, and known failure classification. Keep case-specific prompts and expected sentinels in case YAML automation fields when possible.\n\n## Case Review\n\nFor every changed case:\n\n1. Ensure `steps` describe what to execute, not every command in the underlying implementation.\n2. Ensure `checks` contain observable UI, log, network, or filesystem evidence.\n3. Ensure `diagnostics` are fallback investigation hints, not pass criteria.\n4. Ensure `priority`, `risk`, `ci_eligible`, and `evidence_required` match the actual repeatability and evidence burden.\n5. Put must-have env vars in `env` / `automation_env`; put one-of choices such as URL-or-name in `env_any` / `automation_env_any`.\n6. Ensure linked `skills` and `troubleshooting` ids exist.\n7. Run:\n\n ```bash\n bin/lbs validate\n bin/lbs index --check\n bin/lbs index\n bin/lbs test plan \n ```\n\n## Final Gate\n\nBefore handing off:\n\n- `git diff --stat` should show a focused change set.\n- `skills.index.json` should be regenerated only by `bin/lbs index`.\n- No new asset should contain local credentials, OAuth tokens, API keys, or copied localStorage values.\n- The final note should say which checks ran and which cases or troubleshooting ids changed.\n","skills/skills/langbot-space-ops/SKILL.md":"---\nname: langbot-space-ops\ndescription: Browse and search the LangBot Space marketplaces (plugins, MCP servers, skills) through the Space MCP server. Use when an AI agent needs to discover LangBot extensions on space.langbot.app over MCP. Covers the /mcp endpoint, Personal Access Token (PAT) auth, the tool surface, and client configuration. Triggers on \"langbot space mcp\", \"search langbot plugins\", \"langbot marketplace mcp\", \"space.langbot.app mcp\".\n---\n\n# LangBot Space MCP Operations\n\nLangBot Space (space.langbot.app) exposes an **MCP server** so user-facing AI\nagents can browse and search the marketplaces (plugins, MCP servers, skills) and\nrank live models for automated setup.\n\n## Endpoint\n\n```\nhttps://space.langbot.app/mcp\n```\n\nTransport: **streamable HTTP** (stateless, JSON responses). For a self-hosted\nSpace instance: `http://:8383/mcp`.\n\n## Authentication\n\nReuses the existing **Personal Access Token (PAT)** — the same token the `lbp`\nCLI uses. Create one in your Space account (Profile → Personal Access Tokens),\nthen send it as a Bearer token:\n\n```\nAuthorization: Bearer \n```\n\nRequests without a valid PAT get `401 Unauthorized`.\n\n## Client configuration\n\n```json\n{\n \"mcpServers\": {\n \"langbot-space\": {\n \"url\": \"https://space.langbot.app/mcp\",\n \"headers\": { \"Authorization\": \"Bearer \" }\n }\n }\n}\n```\n\n## Tool surface\n\n| Tool | Purpose |\n| --- | --- |\n| `list_plugins` / `search_plugins` / `get_plugin` | Plugin marketplace |\n| `list_mcp_servers` / `search_mcp_servers` / `get_mcp_server` | MCP-server marketplace |\n| `list_skills` / `search_skills` / `get_skill` | Skill marketplace |\n| `select_models` | Live best-first model list for setup wizards; optional `category` filter |\n\n`list_*` and `search_*` are paged (`page`, `page_size`). `get_*` takes\n`author` + `name`. The tool surface mirrors the REST endpoints under\n`/api/v1/marketplace/*`; `select_models` mirrors `/api/v1/models/selection`.\nAll tools are read-only.\n\n## How to use\n\n1. Create a PAT in your Space account settings.\n2. Point your MCP client at `https://space.langbot.app/mcp` with the Bearer PAT.\n3. Use `search_plugins` / `search_mcp_servers` / `search_skills` to find items,\n then `get_*` for details (e.g. to obtain author/name for installation in\n LangBot itself).\n4. For automatic local-agent setup, call `select_models` (optionally with\n `category`) and choose the first compatible item. Ordering is latest probe\n state (available, unprobed, unavailable), then Space recommendation. Each\n item includes `availability.up`, `last_probed_at`, latency, and HTTP status.\n\n## Runner usage recommendations\n\nUse `search_plugins` with `runner_usage: \"agent\"` for Agent, pipeline, and\nsetup-wizard recommendations, or `runner_usage: \"event\"` for event processors.\nThe component kind remains `Runner`. Only these two exact values are accepted;\nomit the optional field to preserve unfiltered browsing.\n\n```json\n{\"query\":\"\", \"runner_usage\":\"agent\", \"page\":1, \"page_size\":100}\n```\n\nPlugin results include `latest_version` and `runner_usages: string[]`, the\nexplicit union of usages in that latest installable version. Only recommend a\nplugin when this array explicitly contains the target usage. Missing, empty,\nmalformed, or unknown usages must never mean agent-compatible. Event-only\nplugins must never enter Agent recommendations. Empty filtered results are\nvalid while legacy packages await corrected releases; never remove the filter\nto fill a recommendation list.\n\nREST callers use `runner_usage` on both\n`POST /api/v1/marketplace/extensions/search` and the compatibility\n`POST /api/v1/marketplace/plugins/search`; preserve it during fallback. Add\n`\"type_filter\":\"plugin\", \"component_filter\":\"Runner\"` on the unified endpoint.\nUsage is ANDed with other filters before pagination and `total`; MCP/Skill items\ndo not match. Invalid REST values return HTTP 400.\n\nOpen the same filter in the webpage:\n`https://space.langbot.app/market?type=plugin&component=Runner&runner_usage=agent`\n(or `runner_usage=event`). Switch All / Agent / Event in the Runner usage row.\n\n## Implementation & maintenance (for Space developers)\n\n- Server: `internal/controller/mcp/server.go` (official Go MCP SDK\n `github.com/modelcontextprotocol/go-sdk`). Tools call the service layer\n (`PluginService`, `MCPService`, `SkillService`, `ModelStatusService`) directly.\n- Mount: `internal/controller/api.go` at `/mcp` and `/mcp/*any`.\n- Auth: PAT via `AccountService.ValidatePersonalAccessToken`.\n- Docs: `docs/MCP_SERVER.md`.\n\n> When you add, remove, or change a marketplace API endpoint that should be\n> agent-accessible, update the corresponding MCP tool **and** this skill. The\n> MCP tool surface and the API must stay aligned (see `AGENTS.md`).\n\n## Pitfalls\n\n- The PAT prefix is `lbpat_` (Space), distinct from LangBot's `lbk_` API keys.\n- This server is read/browse only; it does not publish or modify marketplace\n items. Use the web UI or REST API (with appropriate auth) for that.\n","skills/skills/langbot-testing/SKILL.md":"","skills/skills/langbot-testing/cases/acp-agent-runner-debug-chat.yaml":"id: acp-agent-runner-debug-chat\ntitle: \"ACP Runner can answer through Debug Chat using real remote Claude\"\nmode: agent-browser\narea: pipeline\ntype: regression\npriority: p2\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - acp\n - claude\n - external-runner\n - pipeline\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nenv_any:\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/pipeline-debug-chat.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\n - LANGBOT_ACP_RUNNER_PIPELINE_URL\n - LANGBOT_ACP_RUNNER_PIPELINE_NAME\n - LANGBOT_E2E_PROMPT\n - LANGBOT_E2E_EXPECTED_TEXT\n - LANGBOT_E2E_EXPECTED_RUNNER_ID\n - LANGBOT_E2E_RESPONSE_TIMEOUT_MS\nautomation_pipeline_url_env: LANGBOT_ACP_RUNNER_PIPELINE_URL\nautomation_pipeline_name_env: LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation_expected_runner_id: \"plugin:langbot-team/ACPRunner/default\"\nautomation_prompt: \"Do not launch any background agent, subagent, task, or worker. In this current ACP session, directly call the MCP tool named langbot_get_current_event exactly once and wait for its result. After it returns, reply exactly ACP_RUNNER_E2E_OK with no other text.\"\nautomation_expected_text: \"ACP_RUNNER_E2E_OK\"\nautomation_response_timeout_ms: \"300000\"\nsetup_automation:\n - \"node:scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env\"\nsetup_provides_env:\n - LANGBOT_ACP_RUNNER_PIPELINE_URL\n - LANGBOT_ACP_RUNNER_PIPELINE_NAME\npreconditions:\n - \"The remote machine has a working Claude Code login and can run npx -y @agentclientprotocol/claude-agent-acp.\"\n - \"LangBot can non-interactively SSH to the remote machine; the runner opens the MCP reverse tunnel automatically.\"\nsteps:\n - \"Open LANGBOT_FRONTEND_URL.\"\n - \"Open the ACP Runner QA pipeline.\"\n - \"Confirm the pipeline AI runner is plugin:langbot-team/ACPRunner/default.\"\n - \"Open Debug Chat.\"\n - \"Ask the real remote Claude ACP agent to call langbot_get_current_event and return ACP_RUNNER_E2E_OK exactly.\"\nchecks:\n - \"UI: Debug Chat shows the user prompt.\"\n - \"UI: Debug Chat shows a Bot response containing ACP_RUNNER_E2E_OK.\"\n - \"Logs: Backend logs include Processing request from person_websocket and Streaming completed for this run.\"\n - \"Logs: No acp runner request error appears for this run.\"\n - \"Console: No unexpected frontend errors appear during Debug Chat.\"\nevidence_required:\n - ui\n - console\n - backend_log\ndiagnostics:\n - \"Use scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env to create/update the pipeline.\"\n - \"For remote Claude on 101, verify ssh yhh@101.34.71.12 can run without password prompts; no separate ssh -R process is required.\"\nsuccess_patterns:\n - \"ACP_RUNNER_E2E_OK\"\n - \"Processing request from person_websocket\"\n - \"Streaming completed\"\nfailure_patterns:\n - \"acp.command_not_found\"\n - \"acp.process_exited\"\n - \"Agent runner plugin:langbot-team/ACPRunner/default execution failed\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - proxy-env-mismatch\n","skills/skills/langbot-testing/cases/agent-run-ledger-audit.yaml":"id: agent-run-ledger-audit\ntitle: \"Persisted Runner run ledger passes end-to-end invariants\"\nmode: probe\narea: agent\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - ledger\n - audit\n - tools\nskills:\n - langbot-testing\nautomation: scripts/e2e/agent-run-ledger-audit.mjs\nsteps:\n - \"Set LANGBOT_AGENT_RUN_ID to audit a specific run, or leave it unset to audit the latest persisted Runner run.\"\n - \"For an external runner's own CLI tools, set LANGBOT_AGENT_TOOL_AUTHORIZATION_MODE=runner-native; keep the default strict mode for Host tool calls.\"\n - \"Read the active LangBot database configuration and inspect the selected run and its ordered events.\"\n - \"Verify completed terminal state, run.completed, paired tool.call.started/completed events, stable tool names, and monotonic ordering.\"\n - \"Compare called tools with the authorization snapshot and validate each advertised tool has owner/source, description, and parameter schema.\"\n - \"Reject malformed persisted event JSON, unrecovered invalid tool arguments, timeout, forbidden, permission-denied, or unauthorized signals.\"\n - \"Record a malformed model tool-argument payload as a recovery warning only when a later tool call succeeds and the run completes.\"\nchecks:\n - \"ledger-audit.json status is pass.\"\n - \"metrics has equal tool_call_started and tool_call_completed counts.\"\n - \"failures is empty and no unauthorized tool call is present.\"\nevidence_required:\n - filesystem\n - metrics\n - api_diagnostic\ndiagnostics:\n - \"Run this immediately after a complex UI Agent task so latest-run selection cannot drift to unrelated traffic.\"\n - \"Use LANGBOT_AGENT_RUN_ID when multiple operators share the test instance.\"\n - \"runner-native mode still checks tool pairing, stable names, ordering, terminal state, and error signals; the external runner's permission system owns its local tools.\"\nsuccess_patterns:\n - \"Agent run ledger audit passed\"\nfailure_patterns:\n - \"tool_call_pairing\"\n - \"forbidden_error_signals\"\n - \"unauthorized_tool_calls\"\ntroubleshooting:\n - box-runtime-silent-action-timeout\n","skills/skills/langbot-testing/cases/agent-runner-async-db-readiness.yaml":"id: agent-runner-async-db-readiness\ntitle: \"Runner async DB readiness probe\"\nmode: probe\narea: release\ntype: smoke\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - async-db\n - aiosqlite\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-async-db-readiness.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-async-db-readiness --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation checks whether a direct aiosqlite in-memory connection can create a table within the readiness timeout.\"\nchecks:\n - \"automation-result.json status is pass or env_issue.\"\n - \"pass means async SQLite tests are worth running.\"\n - \"env_issue means async SQLite-dependent Host pytest probes should be classified as environment-limited until fixed.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"If this probe returns env_issue, run agent-runner-ledger-invariants for fast ledger coverage and skip async ledger pytest as a release blocker in this environment.\"\nsuccess_patterns:\n - \"AIOSQLITE_READY\"\nfailure_patterns:\n - \"aiosqlite readiness timed out\"\ntroubleshooting:\n - aiosqlite-connect-hangs\n","skills/skills/langbot-testing/cases/agent-runner-behavior-matrix.yaml":"id: agent-runner-behavior-matrix\ntitle: \"Runner deterministic behavior matrix probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - deterministic-runner\n - protocol\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-behavior-matrix.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-behavior-matrix --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation reads fixtures/agent-runner/qa-runner-behaviors.json and validates each result sequence through the Host AgentResultNormalizer.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains QA_RUNNER_BEHAVIOR_MATRIX_OK.\"\n - \"The matrix covers ok, stream_ok, empty_output, malformed_result, and controlled_failure.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"fail means the deterministic behavior fixture and Host result normalization disagree.\"\nsuccess_patterns:\n - \"QA_RUNNER_BEHAVIOR_MATRIX_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"RunnerProtocolError\"\n - \"behavior matrix exited\"\n","skills/skills/langbot-testing/cases/agent-runner-fixture-contract.yaml":"id: agent-runner-fixture-contract\ntitle: \"QA Runner fixture contract probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - fixture\n - deterministic-runner\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-fixture-contract.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-fixture-contract --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation imports the QA Runner fixture source and executes normal, streaming, and controlled-failure paths with SDK entities.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains QA_RUNNER_FIXTURE_CONTRACT_OK.\"\n - \"Normal input returns QA_RUNNER_OK:.\"\n - \"Streaming input emits message.delta chunks and completes.\"\n - \"Failure input returns QA_RUNNER_CONTROLLED_FAILURE.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This validates the deterministic fixture source contract. It does not prove the plugin package is installed in a live LangBot instance.\"\nsuccess_patterns:\n - \"QA_RUNNER_FIXTURE_CONTRACT_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"fixture contract exited\"\n","skills/skills/langbot-testing/cases/agent-runner-health-visibility.yaml":"id: agent-runner-health-visibility\ntitle: \"Agent configuration shows whether its selected runner is usable\"\nmode: agent-browser\narea: agent\ntype: feature\npriority: p0\nrisk: medium\nci_eligible: false\ntags:\n - agent-runner\n - agent\n - productization\n - release-gate\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\nautomation: scripts/e2e/agent-runner-health-visibility.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\npreconditions:\n - \"The plugin runtime is enabled and connected, with at least one Runner registered.\"\n - \"The target is a local test instance where a temporary Agent may be created and deleted.\"\nsteps:\n - \"Read the live plugin runtime status and select a registered runner from Agent metadata.\"\n - \"Create a temporary Agent using that runner and open its Runner settings in the WebUI.\"\n - \"Confirm the UI reports that the runner is ready.\"\n - \"Replace the temporary Agent binding with a syntactically valid but unregistered runner id.\"\n - \"Reload Runner settings and confirm the UI reports that the selected runner is unavailable.\"\nchecks:\n - \"UI: A registered runner and connected plugin runtime produce a visible ready status.\"\n - \"UI: A stale runner binding produces a visible unavailable status with recovery actions.\"\n - \"API: Plugin runtime status is connected and Agent metadata contains the registered runner.\"\n - \"Console: No unexpected frontend errors appear during the flow.\"\n - \"Cleanup: The temporary Agent is deleted after evidence is collected.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - api_diagnostic\ndiagnostics:\n - \"A disconnected plugin runtime is env_issue, not a passing unavailable-runner product check.\"\n - \"A ready status inferred only from API metadata without opening the Runner UI is not a pass.\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - proxy-env-mismatch\n","skills/skills/langbot-testing/cases/agent-runner-ledger-concurrency.yaml":"id: agent-runner-ledger-concurrency\ntitle: \"Runner run ledger concurrency and auth pytest probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - pytest\n - run-ledger\n - concurrency\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-concurrency.mjs\npreconditions:\n - \"This Host pytest probe can be slow in the current multi-repo dev environment; keep it in the release gate, but do not treat a timeout as a browser E2E failure without checking pytest logs.\"\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-concurrency --dry-run` first; remove `--dry-run` only after `agent-runner-async-db-readiness` is pass.\"\n - \"Automation resolves LANGBOT_REPO, defaulting to ../LangBot when the env var is unset.\"\n - \"Automation runs selected high-value tests from test_run_ledger_store.py and test_run_ledger_api_auth.py.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"pytest exit status is 0 for selected run ledger claim, lease, status, token, ownership, and active-claim tests.\"\n - \"pytest-stdout.log and pytest-stderr.log are written under LBS_EVIDENCE_DIR.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"env_issue means LANGBOT_REPO/default ../LangBot did not resolve, rtk/uv was unavailable, or the expected test files are missing.\"\n - \"fail means one of the selected LangBot run ledger pytest targets failed or timed out.\"\nsuccess_patterns:\n - \"pytest passed\"\nfailure_patterns:\n - \"pytest exited with status\"\n - \"pytest timed out\"\n - \"Failed to start pytest command\"\ntroubleshooting:\n - aiosqlite-connect-hangs\n","skills/skills/langbot-testing/cases/agent-runner-ledger-contention.yaml":"id: agent-runner-ledger-contention\ntitle: \"Runner ledger SQLite contention probe\"\nmode: probe\narea: release\ntype: regression\npriority: p1\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - stress\n - ledger\n - concurrency\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-contention.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-contention --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation creates 120 queued runs in a file-backed SQLite database and uses eight worker threads to claim runs under write contention.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_CONTENTION_OK.\"\n - \"Every run is claimed once, reaches completed status, and has dispatch_attempts = 1.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe catches obvious exactly-once claim regressions under local SQLite contention; it does not replace async Host pytest or PostgreSQL concurrency checks.\"\nsuccess_patterns:\n - \"LEDGER_CONTENTION_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger contention exited\"\n","skills/skills/langbot-testing/cases/agent-runner-ledger-invariants.yaml":"id: agent-runner-ledger-invariants\ntitle: \"Runner ledger schema and status invariants probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - ledger\n - invariant\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-invariants.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-invariants --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation resolves LANGBOT_REPO, defaulting to the parent LangBot checkout, and imports the sibling SDK from LANGBOT_PLUGIN_SDK_REPO or ../../langbot-plugin-sdk/src.\"\n - \"Automation checks run status sets, terminal status validation, ledger table/index DDL, and a minimal synchronous insert/read path.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_INVARIANTS_OK.\"\n - \"The probe does not require aiosqlite or browser UI.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"env_issue means LANGBOT_REPO/default ../LangBot did not resolve or Python dependencies are unavailable.\"\n - \"fail means a ledger schema/status invariant changed and the release gate needs review.\"\nsuccess_patterns:\n - \"LEDGER_INVARIANTS_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger invariant probe exited\"\n","skills/skills/langbot-testing/cases/agent-runner-ledger-stress.yaml":"id: agent-runner-ledger-stress\ntitle: \"Runner ledger lightweight stress probe\"\nmode: probe\narea: release\ntype: regression\npriority: p1\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - stress\n - ledger\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-stress.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-stress --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation creates 100 queued runs in synchronous SQLite and simulates five runtimes claiming them in priority order.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_STRESS_OK.\"\n - \"Every run is claimed once and reaches a terminal completed status.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe is a fast deterministic stress baseline; it does not replace PostgreSQL/async concurrency tests.\"\nsuccess_patterns:\n - \"LEDGER_STRESS_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger stress exited\"\n","skills/skills/langbot-testing/cases/agent-runner-live-install.yaml":"id: agent-runner-live-install\ntitle: \"QA Runner package installs and registers in LangBot\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - plugin\n - local-install\n - fixture\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_BACKEND_URL\n - LANGBOT_REPO\n - LANGBOT_E2E_LOGIN_USER\nautomation: scripts/e2e/install-qa-plugin-smoke.mjs\nautomation_plugin_package: \"skills/langbot-testing/fixtures/plugins/qa-agent-runner/dist/qa-agent-runner-0.1.0.lbpkg\"\nautomation_expected_plugin_id: \"qa/agent-runner\"\nautomation_expected_tool: \"\"\nautomation_expected_runner_id: \"plugin:qa/agent-runner/default\"\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-live-install --dry-run` first; remove `--dry-run` only after readiness points at a local test LangBot instance.\"\n - \"Automation authenticates the local test user, uploads the QA Runner .lbpkg package, waits for the install task, and reads pipeline metadata.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"/api/v1/plugins lists qa/agent-runner after install.\"\n - \"/api/v1/pipelines/_/metadata lists plugin:qa/agent-runner/default as an available runner.\"\nevidence_required:\n - api_diagnostic\n - filesystem\ndiagnostics:\n - \"This proves the deterministic package installs and registers a runner. It does not prove Debug Chat execution; use a later browser case for that.\"\n - \"If installation fails during dependencies, inspect plugin runtime logs and plugin-dependency-install-offline.\"\nsuccess_patterns:\n - \"qa/agent-runner is installed.\"\nfailure_patterns:\n - \"Plugin install task did not complete successfully\"\n - \"plugin:qa/agent-runner/default is not listed\"\ntroubleshooting:\n - plugin-runtime-timeout\n - plugin-dependency-install-offline\n","skills/skills/langbot-testing/cases/agent-runner-qa-debug-chat.yaml":"id: agent-runner-qa-debug-chat\ntitle: \"QA Runner returns deterministic output through Debug Chat\"\nmode: agent-browser\narea: pipeline\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - pipeline\n - debug-chat\n - fixture\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/pipeline-debug-chat.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation_pipeline_url_env: LANGBOT_QA_RUNNER_PIPELINE_URL\nautomation_pipeline_name_env: LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation_expected_runner_id: \"plugin:qa/agent-runner/default\"\nautomation_prompt: \"hello-live\"\nautomation_expected_text: \"QA_RUNNER_OK:hello-live\"\nautomation_response_timeout_ms: \"120000\"\nautomation_debug_chat_response_p95_ms: \"120000\"\nautomation_reset_debug_chat: \"1\"\nsetup_automation:\n - \"case:agent-runner-live-install\"\n - \"node:scripts/e2e/ensure-qa-agent-runner-pipeline.mjs --write-env\"\nsetup_provides_env:\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nsteps:\n - \"Open LANGBOT_FRONTEND_URL.\"\n - \"Open the pipeline from LANGBOT_QA_RUNNER_PIPELINE_URL or LANGBOT_QA_RUNNER_PIPELINE_NAME.\"\n - \"Confirm the pipeline AI runner is plugin:qa/agent-runner/default.\"\n - \"Open Debug Chat.\"\n - \"Send: hello-live.\"\nchecks:\n - \"UI: The user message appears in Debug Chat.\"\n - \"UI: A Bot message appears and contains QA_RUNNER_OK:hello-live.\"\n - \"API diagnostic: pipeline config uses plugin:qa/agent-runner/default.\"\n - \"Console: No unexpected frontend runtime errors appear during the send/receive path.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - network\n - api_diagnostic\ndiagnostics:\n - \"This is the deterministic live execution proof that sits after fixture contract and live install.\"\n - \"If the runner id mismatch is reported, rerun ensure-qa-agent-runner-pipeline.mjs --write-env.\"\nsuccess_patterns:\n - \"QA_RUNNER_OK:hello-live\"\nfailure_patterns:\n - \"plugin:qa/agent-runner/default execution failed\"\n - \"Action invoke_llm_stream call timed out\"\n - \"Agent runner temporarily unavailable\"\ntroubleshooting:\n - plugin-runtime-timeout\n","skills/skills/langbot-testing/cases/agent-runner-release-preflight.yaml":"id: agent-runner-release-preflight\ntitle: \"Agent runner release gate preflight validates environment readiness\"\nmode: agent-browser\narea: release\ntype: smoke\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - release-gate\n - preflight\n - environment\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nenv_any:\n - LANGBOT_LOCAL_AGENT_PIPELINE_URL|LANGBOT_LOCAL_AGENT_PIPELINE_NAME\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/agent-runner-release-preflight.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\nautomation_env_any:\n - LANGBOT_LOCAL_AGENT_PIPELINE_URL|LANGBOT_LOCAL_AGENT_PIPELINE_NAME\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\npreconditions:\n - \"LANGBOT_LOCAL_AGENT_PIPELINE_URL or LANGBOT_LOCAL_AGENT_PIPELINE_NAME points to the local-agent release pipeline.\"\n - \"LANGBOT_ACP_RUNNER_PIPELINE_URL or LANGBOT_ACP_RUNNER_PIPELINE_NAME points to the ACP Runner release pipeline.\"\n - \"The active browser profile is authenticated for the same LangBot backend.\"\n - \"By default the preflight performs a cheap model test for the local-agent primary model; set LANGBOT_PREFLIGHT_TEST_MODELS=0 only when deliberately classifying model credentials outside this run.\"\nsteps:\n - \"Open LANGBOT_FRONTEND_URL with the configured browser profile.\"\n - \"Use the browser token to call LangBot backend readiness APIs without printing token values.\"\n - \"Check plugin runtime status, Box status, required runner plugins, qa-plugin-smoke, and qa_plugin_echo.\"\n - \"Resolve the local-agent and ACP Runner QA pipelines from their case-specific env vars.\"\n - \"Assert each pipeline uses the expected runner id.\"\n - \"Assert the external runner pipeline uses the expected runner id.\"\n - \"Assert the local-agent primary model advertises func_call and vision for the full release gate.\"\n - \"Run the local-agent primary model test endpoint unless LANGBOT_PREFLIGHT_TEST_MODELS=0.\"\nchecks:\n - \"API diagnostic: api-diagnostic.json has no blockers and no env_issues.\"\n - \"API diagnostic: required pipelines resolve to plugin:langbot-team/LocalAgent/default and plugin:langbot-team/ACPRunner/default.\"\n - \"API diagnostic: qa_plugin_echo is exposed by /api/v1/tools.\"\n - \"API diagnostic: local-agent model check catches invalid credentials or missing func_call/vision before release E2E starts.\"\n - \"Secret safety: token values, api keys, and provider secrets are not printed.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - network\n - api_diagnostic\ndiagnostics:\n - \"blocked means the test instance is not configured for the full release gate: missing pipeline, wrong runner id, or missing plugin.\"\n - \"env_issue means the runtime or upstream dependency is not usable: backend unavailable, plugin runtime down, Box down, or the local-agent model cannot pass a model test.\"\n - \"If qa_mcp_echo is absent here, continue to mcp-stdio-register before mcp-stdio-tool-call; qa_mcp_echo is not required before registration.\"\n - \"If the model check fails with invalid api key, switch the local-agent release pipeline to a known-good func_call model before diagnosing runner code.\"\nsuccess_patterns:\n - \"Release gate preflight passed\"\nfailure_patterns:\n - \"Preflight blocked\"\n - \"Preflight environment issue\"\n - \"invalid api key\"\n - \"runner.llm_error\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - local-agent-model-route-unavailable\n - proxy-env-mismatch\n","skills/skills/langbot-testing/cases/agent-runner-runtime-chaos.yaml":"id: agent-runner-runtime-chaos\ntitle: \"Runner SDK runtime chaos pytest probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - pytest\n - runtime\n - sdk\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-runtime-chaos.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-runtime-chaos --dry-run` first; remove `--dry-run` after checking the SDK repo target and evidence directory.\"\n - \"Automation resolves LANGBOT_PLUGIN_SDK_REPO, defaulting to ../../langbot-plugin-sdk when the env var is unset.\"\n - \"Automation runs the existing SDK pytest files tests/runtime/plugin/test_mgr_runner.py and tests/runtime/test_pull_api_handlers.py.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"pytest exit status is 0 for the existing Runner runtime and pull API handler tests.\"\n - \"pytest-stdout.log and pytest-stderr.log are written under LBS_EVIDENCE_DIR.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe does not open the WebUI; it runs SDK pytest targets directly.\"\n - \"env_issue means LANGBOT_PLUGIN_SDK_REPO/default ../../langbot-plugin-sdk did not resolve, rtk/uv was unavailable, or the expected test files are missing.\"\n - \"fail means the existing SDK runtime pytest target failed or timed out.\"\nsuccess_patterns:\n - \"pytest passed\"\nfailure_patterns:\n - \"pytest exited with status\"\n - \"pytest timed out\"\n - \"Failed to start pytest command\"\n","skills/skills/langbot-testing/cases/bot-event-routing-product-flow.yaml":"","skills/skills/langbot-testing/cases/box-mcp-heartbeat-recovery.yaml":"","skills/skills/langbot-testing/cases/claude-code-agent-debug-chat.yaml":"","skills/skills/langbot-testing/cases/codex-agent-debug-chat.yaml":"","skills/skills/langbot-testing/cases/dify-agent-debug-chat.yaml":"","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-cross-pipeline-isolation.yaml":"","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-fault-recovery.yaml":"","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-load.yaml":"","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-slow-load.yaml":"","skills/skills/langbot-testing/cases/langbot-fault-taxonomy-contract.yaml":"","skills/skills/langbot-testing/cases/langbot-live-backend-latency.yaml":"","skills/skills/langbot-testing/cases/langbot-live-backend-log-health.yaml":"","skills/skills/langbot-testing/cases/langbot-live-control-plane-api.yaml":"","skills/skills/langbot-testing/cases/langbot-overhead-accounting-contract.yaml":"","skills/skills/langbot-testing/cases/langbot-space-debug-chat-concurrency-smoke.yaml":"","skills/skills/langbot-testing/cases/langrag-kb-retrieve.yaml":"","skills/skills/langbot-testing/cases/langrag-parser-golden-e2e.yaml":"","skills/skills/langbot-testing/cases/langrag-sentinel-kb-discover.yaml":"","skills/skills/langbot-testing/cases/local-agent-basic-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-combo-rag-compaction-tool-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-complex-coding-task-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-context-compaction-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-effective-prompt-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-model-fallback-before-first-chunk-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-multimodal-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-multitool-rag-compaction-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-nonstreaming-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-parallel-tools-rag-compaction-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-plugin-tool-call-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-rag-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-rag-multimodal-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-steering-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-streaming-post-commit-failure-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-tool-error-recovery-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-tool-loop-limit-debug-chat.yaml":"","skills/skills/langbot-testing/cases/mcp-stdio-register.yaml":"","skills/skills/langbot-testing/cases/mcp-stdio-tool-call.yaml":"","skills/skills/langbot-testing/cases/pipeline-debug-chat-performance.yaml":"","skills/skills/langbot-testing/cases/pipeline-debug-chat.yaml":"","skills/skills/langbot-testing/cases/plugin-e2e-smoke.yaml":"","skills/skills/langbot-testing/cases/provider-deepseek.yaml":"","skills/skills/langbot-testing/cases/qa-plugin-smoke-live-install.yaml":"","skills/skills/langbot-testing/cases/sandbox-skill-authoring-e2e.yaml":"","skills/skills/langbot-testing/cases/sandbox-skill-authoring-edit-existing-e2e.yaml":"","skills/skills/langbot-testing/cases/skill-discovery-via-mcp-gateway.yaml":"","skills/skills/langbot-testing/cases/webui-login-state.yaml":"","skills/skills/langbot-testing/cases/wizard-onebot-agent-runtime.yaml":"","skills/skills/langbot-testing/cases/wizard-runner-marketplace-catalog.yaml":"","skills/skills/langbot-testing/cases/wizard-scenario-routing.yaml":"","skills/skills/langbot-testing/cases/workspace-compatibility-preflight.yaml":"","skills/skills/langbot-testing/cases/workspace-plugin-pages-smoke.yaml":"","skills/skills/langbot-testing/cases/workspace-repository-contracts.yaml":"","skills/skills/langbot-testing/references/agent-runner-qa-workflow.md":"","skills/skills/langbot-testing/references/agent-runner-release-gate.md":"","skills/skills/langbot-testing/references/dify-agent-runner.md":"","skills/skills/langbot-testing/references/langrag-knowledge-base.md":"","skills/skills/langbot-testing/references/local-agent-runner-coverage.md":"","skills/skills/langbot-testing/references/local-agent-runner.md":"","skills/skills/langbot-testing/references/mcp-stdio-testing.md":"","skills/skills/langbot-testing/references/model-provider-testing.md":"","skills/skills/langbot-testing/references/performance-reliability-testing.md":"","skills/skills/langbot-testing/references/pipeline-debug-chat.md":"","skills/skills/langbot-testing/references/plugin-e2e-smoke.md":"","skills/skills/langbot-testing/references/sandbox-skill-authoring.md":"","skills/skills/langbot-testing/references/skill-all-tool-acceptance.md":"","skills/skills/langbot-testing/references/troubleshooting.md":"","skills/skills/langbot-testing/references/web-ui-testing.md":"","skills/skills/langbot-testing/references/workspace-release-testing.md":"","skills/skills/langbot-testing/suites/agent-runner-release-gate.yaml":"","skills/skills/langbot-testing/suites/core-smoke.yaml":"","skills/skills/langbot-testing/suites/eba-release-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-debug-chat-isolation-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-debug-chat-load-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-live-backend-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-performance-contract-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-performance-reliability-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-user-path-performance-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-workspace-contract-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-workspace-release-gate.yaml":"","skills/skills/langbot-testing/suites/local-agent-gate.yaml":"","skills/skills/langbot-testing/troubleshooting/agent-runner-actor-context-fields.yaml":"","skills/skills/langbot-testing/troubleshooting/aiosqlite-connect-hangs.yaml":"","skills/skills/langbot-testing/troubleshooting/ambiguous-runner-default-label.yaml":"","skills/skills/langbot-testing/troubleshooting/backend-not-listening.yaml":"","skills/skills/langbot-testing/troubleshooting/box-runtime-silent-action-timeout.yaml":"","skills/skills/langbot-testing/troubleshooting/box-session-conflict-logical-metadata.yaml":"","skills/skills/langbot-testing/troubleshooting/debug-chat-history-contaminates-automation.yaml":"","skills/skills/langbot-testing/troubleshooting/dynamic-form-missing-config-id.yaml":"","skills/skills/langbot-testing/troubleshooting/e2b-extra-mount-sync-missing.yaml":"","skills/skills/langbot-testing/troubleshooting/local-agent-mcp-text-result-empty-follow-up.yaml":"","skills/skills/langbot-testing/troubleshooting/local-agent-model-route-unavailable.yaml":"","skills/skills/langbot-testing/troubleshooting/marketplace-network-flaky.yaml":"","skills/skills/langbot-testing/troubleshooting/mcp-stdio-args-not-applied.yaml":"","skills/skills/langbot-testing/troubleshooting/nsjail-cli-compatibility.yaml":"","skills/skills/langbot-testing/troubleshooting/pipeline-form-controlled-warning.yaml":"","skills/skills/langbot-testing/troubleshooting/plugin-dependency-install-offline.yaml":"","skills/skills/langbot-testing/troubleshooting/plugin-runtime-timeout.yaml":"","skills/skills/langbot-testing/troubleshooting/provider-image-parse-error.yaml":"","skills/skills/langbot-testing/troubleshooting/proxy-env-mismatch.yaml":"","skills/skills/langbot-testing/troubleshooting/sandbox-native-tools-unavailable.yaml":"","skills/skills/langbot-testing/troubleshooting/socks-proxy-without-socksio.yaml":"","skills/skills/langbot-testing/troubleshooting/survey-widget-blocks-debug-chat.yaml":"","skills/skills/langbot-testing/troubleshooting/telemetry-proxy-noise.yaml":"","skills/skills/langbot-testing/troubleshooting/tool-name-collision-between-mcp-and-plugin.yaml":"","skills/skills/langbot-testing/troubleshooting/uv-run-resyncs-local-sdk.yaml":""},"files":{"AGENTS.md":"# AGENTS.md\n\nThis file guides code agents working in the LangBot main repository. `CLAUDE.md` is a symlink to this file.\n\nRead `ARCHITECTURE.md` before non-trivial backend, frontend, runtime, plugin, Box, MCP, persistence, or cross-repo SDK changes. This file is the working checklist; `ARCHITECTURE.md` is the system map.\n\n## Quick Facts\n\n- Python backend: `>=3.11,<4.0`, dependencies managed by `uv`.\n- Frontend: `web/` is Vite + React Router 7 + shadcn/ui + Tailwind, managed by `pnpm`.\n- Backend framework: Quart served by Hypercorn on `api.port`, default `5300`.\n- Frontend dev server: `web/` on `3000`, with `VITE_API_BASE_URL` pointing at the backend.\n- Plugin/Box/runtime contracts live in sibling repo `langbot-plugin-sdk`, pinned as `langbot-plugin` in `pyproject.toml`.\n\n## Essential Commands\n\n```bash\nuv sync --dev\nuv run main.py\nuv run pre-commit install\n\ncd web\npnpm install\npnpm dev\npnpm build\n```\n\nUseful focused tests:\n\n```bash\nuv run pytest tests/unit_tests -q\nuv run pytest tests/integration -q\nuv run pytest tests/integration/persistence -q\nuv run pytest tests/manual/mcp_smoke.py\n\ncd web\npnpm lint\npnpm test:e2e\n```\n\nRun the narrowest useful test first, then broader checks when confidence is needed.\n\n## Where to Look\n\n- Architecture map: `ARCHITECTURE.md`.\n- Dev environment guide: https://langbot.app/docs/zh/develop/dev-config.\n- Plugin runtime / CLI / SDK debugging: https://langbot.app/docs/zh/develop/plugin-runtime.\n- API-key auth: `docs/API_KEY_AUTH.md`.\n- Service API CLI: [`langbot-cli`](https://github.com/langbot-app/langbot-cli) (`lbctl`) manages running Workspaces; it is separate from the SDK's `lbp` CLI.\n- Box deep-dive notes: `docs/review/box-architecture.md` and related files.\n- In-repo skills: `skills/` is the single source of truth for LangBot agent skills.\n- SDK repo: `../langbot-plugin-sdk/` when changing shared entities, plugin APIs, action protocol, `lbp rt`, or `lbp box`.\n\n## Cross-Repo SDK Work\n\nWhen changing SDK contracts used by LangBot:\n\n```bash\n# from langbot-plugin-sdk, with LangBot's .venv active\nuv pip install .\n\n# from LangBot, preserve the locally installed SDK\nuv run --no-sync main.py\n```\n\nFor standalone runtime debugging:\n\n```bash\n# in langbot-plugin-sdk\nuv run --no-sync lbp rt\nuv run --no-sync lbp box\n\n# in LangBot\nuv run --no-sync main.py --standalone-runtime\nuv run --no-sync main.py --standalone-box\n```\n\nConfig keys to verify in `data/config.yaml` / `src/langbot/templates/config.yaml`:\n\n- Plugin runtime: `plugin.runtime_ws_url`, default Docker host `langbot_plugin_runtime:5400/control/ws`.\n- Box runtime: `box.enabled`, `box.backend`, `box.runtime.endpoint`, Docker host `langbot_box:5410`.\n- API/MCP auth: `api.global_api_key`.\n\n## Change Rules\n\n- HTTP API changes that should be agent-accessible must update the matching MCP tool in `src/langbot/pkg/api/mcp/server.py` and the relevant skill under `skills/` in the same pass.\n- When changing Service API routes or capabilities used by `lbctl`, check compatibility with the separate `langbot-cli` repository.\n- New schema changes use Alembic under `src/langbot/pkg/persistence/alembic/versions/`. LangBot 4.x does not support upgrading 3.x databases.\n- New platform behavior belongs in platform adapters only for platform translation; pipeline/business logic belongs in `pkg/pipeline/` or services.\n- User-facing strings must support i18n; follow the backend language rules below for server-generated content.\n- Conversation payload columns, including new ones, must use `MediaReferenceText` and persist image references only. See the [media persistence contract](ARCHITECTURE.md#persistence-and-configuration).\n- Interactive session lists use `SessionSummaryReader` to read only `monitoring_sessions`; `SessionSummaryWriter` maintains summaries in the source transaction. Historical initialization completes before writers start, never in requests or application background tasks. Follow the [conversation summary contract](ARCHITECTURE.md#conversation-summary-boundary) and keep `uv run pytest --required-suite session_summary` passing.\n- Code comments and docstrings must be English.\n- Keep compatibility and security in mind; LangBot is used in both self-hosted/community and toB deployments.\n- Commit message format: `(): `.\n\n## Backend Message Language\n\n- LangBot-owned business errors, runtime notices, and default messages sent to users must use the owning Workspace's message language. Keep this independent of each member's WebUI language; an unset Workspace language defaults to `en_US`.\n- Reuse `src/langbot/pkg/i18n/` and the existing `message_language` entry in `WorkspaceMetadata`. Use `t()` with stable English templates and named parameters; add matching keys and placeholders to all eight catalogs: `en_US`, `zh_Hans`, `zh_Hant`, `ja_JP`, `es_ES`, `ru_RU`, `th_TH`, and `vi_VN`. Avoid new translation frameworks or duplicate language settings.\n- Resolve language from a trusted Workspace or bot execution context, and isolate it across concurrent requests, tasks, and adapter callbacks. Browser language hints may supply account-only or unauthenticated response language, but must not override a resolved Workspace or authorize tenant selection. Reading/changing the setting requires `workspace.view` / `workspace.update`.\n- Pipeline notices should be configurable in Pipeline output settings through `pipeline_notice()`. Blank overrides inherit the Workspace translation. Preserve user-authored text literally except for documented placeholders, and retain existing visibility switches. Do not rewrite historical messages or custom text when the language changes.\n- China-only adapters and configuration/field validation diagnostics do not require localization. An adapter also serving international users is not exempt merely because it supports a Chinese platform.\n- Preserve raw provider/plugin diagnostics, tracebacks, protocol values, and machine-readable error/status codes. Translate LangBot's own category or explanation around them; do not translate arbitrary errors with substring replacement. Plugin-authored messages remain the plugin's localization responsibility.\n- Keep changes small and maintainable. Reuse existing persistence and response boundaries; delete obsolete fields only after verifying they have no remaining readers, including compatibility paths.\n- When changing language behavior, verify catalog/placeholder parity, Workspace isolation and permissions, and Pipeline inheritance/custom overrides. Update affected configuration snapshots and API/MCP documentation. Implementation details and focused test locations: [Backend message language](docs/architecture/backend-i18n.md).\n\n## Runtime Pitfalls\n\n- Local stdio Plugin Runtime disconnects do not auto-reconnect; restart LangBot if that path breaks.\n- Orphan runtime processes on `5400`/`5401` commonly break plugin debugging.\n- Use `uv run --no-sync` after locally installing the SDK, or `uv` may restore the pinned package.\n- A false Box “no backend” often means Docker is running but the current user lacks Docker socket permission.\n- Do not confuse external MCP servers LangBot connects to (`pkg/provider/tools/loaders/mcp.py`) with LangBot's own `/mcp` server (`pkg/api/mcp/`).\n- `CLAUDE.md` is a symlink to this file; edit `AGENTS.md`, not the symlink.\n\n## Principles\n\n- Keep it simple, stupid.\n- Entities should not be multiplied unnecessarily.\n- 八荣八耻\n\n 以瞎猜接口为耻，以认真查询为荣。\n 以模糊执行为耻，以寻求确认为荣。\n 以臆想业务为耻，以人类确认为荣。\n 以创造接口为耻，以复用现有为荣。\n 以跳过验证为耻，以主动测试为荣。\n 以破坏架构为耻，以遵循规范为荣。\n 以假装理解为耻，以诚实无知为荣。\n 以盲目修改为耻，以谨慎重构为荣。\n","CLAUDE.md":"AGENTS.md","skills/AGENTS.md":"# Agent Workflow\n\nThis repository stores reusable LangBot agent-testing assets. Keep changes structured so the next agent does not need to rediscover paths.\n\n## First Steps\n\n1. Read `skills/.env` before using local URLs, paths, browser profiles, or proxy defaults. If present, `skills/.env.local` overrides it for this machine and must not be committed. On a new machine, copy `skills/.env.example` to `skills/.env.local` first.\n2. Pick the smallest relevant skill:\n - `langbot-env-setup` for environment, browser, OAuth, proxy, and startup.\n - `langbot-testing` for WebUI, provider, pipeline, cases, and troubleshooting.\n - `langbot-skills-maintenance` for adding, deduplicating, or auditing this skills repository.\n3. Prefer existing cases and troubleshooting entries before exploring from scratch.\n\n## Editing Rules\n\n- UI/browser testing is the primary QA path. API/curl checks are diagnostic only and cannot make a UI case pass by themselves.\n- Put skills under `skills//`.\n- Keep `SKILL.md` concise; move detailed workflows to `references/`.\n- Put reusable test paths in `cases/*.yaml`.\n- New or edited cases must include `priority`, `risk`, `ci_eligible`, and `evidence_required` so agents can select the right test set without rereading every file.\n- Use `env_any` / `automation_env_any` for one-of machine inputs, such as `LANGBOT_PIPELINE_URL|LANGBOT_PIPELINE_NAME`; do not list those alternatives as separate all-required env keys.\n- Put reusable groups of cases in `suites/*.yaml` rather than hardcoding test sets in docs or CLI code.\n- Put growing failure knowledge in `troubleshooting/*.yaml`.\n- Do not hardcode local ports in testing docs; use `skills/.env` variables and machine-local `skills/.env.local` overrides.\n- Do not store secrets, API keys, OAuth tokens, or localStorage token values.\n\n## Required Checks\n\nAfter structural changes, run:\n\n```bash\nbin/lbs validate\n```\n\nAfter changing skills, cases, or troubleshooting assets, run:\n\n```bash\nbin/lbs index\n```\n\nUse `bin/lbs env show` to inspect defaults and `bin/lbs env doctor` when diagnosing local environment readiness. Env output is redacted by default; do not work around that by printing raw secrets.\n`bin/lbs` is a generated local wrapper. If it is missing on a fresh checkout, run `npm run bootstrap` from this directory first; `npm install` also regenerates it via `prepare`.\nUse `bin/lbs fixture check` before fixture-heavy cases such as MCP, RAG, multimodal, or plugin smoke tests.\nUse `bin/lbs case list --ready` for cases that have no missing machine inputs and no manual preconditions. Use `bin/lbs case list --machine-ready` when you want to keep `manual-check` candidates and confirm their preconditions yourself.\n\nBefore executing a saved QA path, generate the agent-facing plan:\n\n```bash\nbin/lbs test plan \n```\n\nRead the plan readiness sections before running the browser path. Missing env,\nautomation env, or fixture readiness means the case is not ready to execute and\nshould be marked `blocked` or fixed first.\n`manual_check` means machine inputs are present but the agent must verify the\ndeclared `preconditions` or `setup` items before executing the UI path. Do not\nturn a `manual_check` case into `pass` until those items were checked in the\nsame run.\n\nBefore executing a group of saved QA paths, generate the suite plan:\n\n```bash\nbin/lbs suite plan \n```\n\nUse `bin/lbs suite start ` to create a shared suite run id, suite evidence root, per-case evidence directories, and `suite-start.json`/`suite-start.md` handoff files. Then run `bin/lbs suite report --evidence-dir ` to aggregate case results.\nAutomation scripts write `automation-result.json`; write the final per-case `result.json` with `bin/lbs test result --result --reason --evidence-dir --evidence ` after collecting the required evidence. A `pass` result must include all required evidence.\nFor runner-specific Debug Chat cases, prefer case-specific pipeline env keys such as `LANGBOT_LOCAL_AGENT_PIPELINE_URL` over the generic `LANGBOT_PIPELINE_URL`; otherwise an agent can accidentally test the wrong runner.\n","skills/README.md":"# LangBot Skills\n\nThis directory is the **single source of truth** for LangBot's agent skills —\nreusable, on-demand instruction packs for AI agents (Claude Code, Codex, Cursor,\nand LangBot's own Local Agent) working with the LangBot ecosystem.\n\n> These skills were consolidated here from the former `langbot-app/langbot-skills`\n> repository (now archived). Documentation and the landing page link here; do not\n> re-copy skill content elsewhere — link to this directory instead.\n\n## Skill catalog\n\n| Skill | What it covers |\n| --- | --- |\n| [`langbot-dev`](skills/langbot-dev) | Core backend + web frontend development (Quart, Vite, API, migrations, MCP server) |\n| [`langbot-plugin-dev`](skills/langbot-plugin-dev) | Plugin SDK / component development, debugging, WebSocket testing |\n| [`langbot-deploy`](skills/langbot-deploy) | Docker / Compose / Kubernetes deployment, config.yaml, Box runtime, global API key |\n| [`langbot-testing`](skills/langbot-testing) | WebUI / e2e QA harness, cases, fixtures, troubleshooting (the `bin/lbs` CLI) |\n| [`langbot-env-setup`](skills/langbot-env-setup) | Local dev/test environment, browser access, OAuth, proxy, startup |\n| [`langbot-mcp-ops`](skills/langbot-mcp-ops) | Operating a LangBot instance through its MCP server (`/mcp`) |\n| [`langbot-space-ops`](skills/langbot-space-ops) | Browsing the LangBot Space marketplaces through the Space MCP server |\n| [`langbot-eba-adapter-dev`](skills/langbot-eba-adapter-dev) | Building platform adapters for the Event-Based Agents architecture |\n| [`langbot-skills-maintenance`](skills/langbot-skills-maintenance) | Adding, deduplicating, and auditing skills in this directory |\n\n`skills.index.json` is the machine-readable index (regenerate with `bin/lbs index`).\n\n## Quick start (for an AI agent)\n\n1. Read this README, `AGENTS.md`, and `docs/user-guide.md` to understand the layout.\n2. Read `skills/.env` for shared local defaults. On a new machine, copy\n `skills/.env.example` to `skills/.env.local` (gitignored) and override\n machine-specific values there. Never commit secrets.\n3. Pick the smallest relevant skill from the catalog above and follow its\n `SKILL.md`.\n\n## The `lbs` CLI\n\nThe testing assets ship with a small CLI (`bin/lbs`, Node >= 22.6). The\n`bin/lbs` wrapper is a generated local entrypoint; on a fresh checkout, run\n`npm run bootstrap` once if it is missing. `npm install` also regenerates it via\nthe `prepare` script.\n\n```bash\nnpm run bootstrap # create bin/lbs if missing\nbin/lbs validate # validate skills/cases/troubleshooting structure\nbin/lbs index # regenerate skills.index.json\nbin/lbs env show # inspect resolved env defaults (redacted)\nbin/lbs env doctor # diagnose local environment readiness\nbin/lbs case list --ready\nbin/lbs test plan \nbin/lbs suite plan langbot-debug-chat-load-gate\n```\n\n## Maintenance rule\n\nWhen the LangBot / LangBot Space **API or MCP server changes**, the\ncorresponding skill here MUST be updated in the same change. The MCP tool\nsurface, the API, and these skills are kept in lockstep — see each repo's\n`AGENTS.md`.\n","skills/docs/user-guide.md":"# LangBot QA Skills User Guide\n\nUse this guide as the first operational path after reading `README.md` and\n`AGENTS.md`.\n\n## 1. Configure Local Inputs\n\nRead `skills/.env`, then create `skills/.env.local` for machine-local values.\nDo not commit `.env.local`, browser profiles, reports, tokens, API keys, OAuth\nstate, or provider credentials.\n\nMinimum local fields for live browser QA:\n\n```bash\nLANGBOT_REPO=/path/to/LangBot\nLANGBOT_WEB_REPO=/path/to/LangBot/web\nLANGBOT_BACKEND_URL=http://127.0.0.1:5300\nLANGBOT_FRONTEND_URL=http://127.0.0.1:3000\nLANGBOT_DEV_FRONTEND_URL=http://127.0.0.1:3000\nLANGBOT_BROWSER_PROFILE=/path/to/langbot-browser-profile\nLANGBOT_CHROMIUM_EXECUTABLE=/path/to/chromium-or-playwright-chrome\nLANGBOT_E2E_LOGIN_USER=qa-local@example.com\n```\n\n`LANGBOT_E2E_LOGIN_USER` is a local QA account. The setup automation uses the\nLangBot recovery key from the active checkout to initialize or refresh that\nlocal account and write a browser `localStorage` token. It does not need the\nuser's GitHub or Space credentials.\n\n## 2. Check Readiness\n\nFrom `skills/`:\n\n```bash\nbin/lbs env show\nbin/lbs env doctor\nbin/lbs validate\nbin/lbs index --check\n```\n\n`env doctor` should report reachable backend and frontend URLs before live\nbrowser cases are run. Missing Space provider credentials are not a LangBot\nproduct pass; classify them as `env_issue` and configure the local Space\nprovider before measuring Debug Chat performance.\n\n## 3. Start Services\n\nStart the backend from `LANGBOT_REPO`:\n\n```bash\ncd \"$LANGBOT_REPO\"\nuv run main.py\n```\n\nStart the standalone frontend from `LANGBOT_WEB_REPO` and point it at the\nbackend:\n\n```bash\ncd \"$LANGBOT_WEB_REPO\"\nVITE_API_BASE_URL=\"$LANGBOT_BACKEND_URL\" pnpm dev --host 0.0.0.0\n```\n\nIf `VITE_API_BASE_URL` is missing, browser tests can load the Vite page but send\nAPI requests to the frontend port, which produces false UI failures.\n\n## 4. Prepare User-Path Fixtures\n\nFor local-agent Debug Chat cases and the user-path performance gate:\n\n```bash\nnode scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\n```\n\nThe script:\n\n- refreshes the local QA login and browser token;\n- marks the local wizard as skipped;\n- creates or updates a local QA pipeline;\n- scans Space LLM models, tests candidates, and switches to the first working\n Space model with tested fallback models;\n- writes `LANGBOT_PIPELINE_URL`, `LANGBOT_PIPELINE_NAME`, and local-agent\n pipeline/model variables into `skills/.env.local`;\n- returns `env_issue` when no Space model can be scanned or tested.\n\nUseful model controls:\n\n```bash\nLANGBOT_E2E_MODEL_TEST_LIMIT=8\nLANGBOT_E2E_MODEL_FALLBACK_COUNT=3\nLANGBOT_E2E_SKIP_MODEL_UUIDS=uuid-a,uuid-b\nLANGBOT_E2E_SKIP_MODEL_NAMES=model-a,model-b\nLANGBOT_E2E_SCAN_SPACE_MODELS=true\n```\n\nThe setup writes a current-runtime compatibility `max-round` value into the\npipeline config because this backend still reads that field directly during\nmessage truncation. Do not treat it as a long-term QA contract.\n\n## 5. Run Gates\n\nCross-repository contract gate (no model calls):\n\n```bash\nbin/lbs suite plan langbot-workspace-contract-gate\nbin/lbs suite run langbot-workspace-contract-gate --run-id langbot-workspace-contract-local\n```\n\nTop-down workspace release gate (browser workflows plus one complex Agent task):\n\n```bash\nbin/lbs suite plan langbot-workspace-release-gate\nbin/lbs suite run langbot-workspace-release-gate --run-id langbot-workspace-release-local --include-manual-check\n```\n\nRun `agent-run-ledger-audit` immediately after the complex Agent case. Set\n`LANGBOT_AGENT_RUN_ID` when the test instance has concurrent operator traffic.\n\nFast contract gate, no live service required:\n\n```bash\nbin/lbs suite run langbot-performance-contract-gate --run-id langbot-contract-local\n```\n\nLive backend gate:\n\n```bash\nbin/lbs suite run langbot-live-backend-gate --run-id langbot-backend-local\n```\n\nBrowser-visible user-path performance gate:\n\n```bash\nbin/lbs suite plan langbot-user-path-performance-gate\nbin/lbs suite run langbot-user-path-performance-gate --run-id langbot-user-path-local --include-manual-check\n```\n\nControlled Debug Chat message-path load gate (manual/non-required; run fake-provider cases serially when they share `LANGBOT_FAKE_PROVIDER_URL`):\n\n```bash\nbin/lbs suite plan langbot-debug-chat-load-gate\nbin/lbs test run langbot-fake-provider-debug-chat-load --run-id langbot-fake-load-local\nbin/lbs test run langbot-fake-provider-debug-chat-slow-load --run-id langbot-fake-slow-local\nbin/lbs test run langbot-fake-provider-debug-chat-fault-recovery --run-id langbot-fake-fault-local\nbin/lbs test run langbot-space-debug-chat-concurrency-smoke --run-id langbot-space-smoke-local\n```\n\nCross-pipeline Debug Chat isolation is a separate manual regression gate because\ncurrent releases may fail it due to product bug #2286:\n\n```bash\nbin/lbs suite plan langbot-debug-chat-isolation-gate\nbin/lbs suite run langbot-debug-chat-isolation-gate --run-id langbot-debug-chat-isolation-local --include-manual-check\n```\n\nStart with `langbot-fake-provider-debug-chat-load`. It launches a local\nOpenAI-compatible fake provider, creates the matching provider/model/pipeline,\nthen sends concurrent WebSocket Debug Chat messages through the real backend.\nUse `langbot-fake-provider-debug-chat-slow-load` to measure the same path under\ndeterministic streaming latency. Use\n`langbot-fake-provider-debug-chat-fault-recovery` to inject bounded provider\nHTTP failures and confirm later Debug Chat requests recover. Use the separate\n`langbot-debug-chat-isolation-gate` to verify that concurrent Debug Chat traffic\non two pipelines does not leak assistant responses across pipeline boundaries;\ncurrent releases may fail that gate because of #2286, so keep it out of the\nnormal load gate until the product fix lands.\nUse `langbot-space-debug-chat-concurrency-smoke` only as a low-volume live\nprovider smoke; it includes Space/model/network latency and should be compared\nagainst the fake-provider baseline before attributing failures to LangBot.\n\n`manual_check` means the agent must confirm the declared preconditions for that\nrun window. When setup automation is declared, run output may stop early with\n`env_issue`; fix that environment input before treating the product path as\nmeasured.\n\n## 6. Read Results\n\nSuite reports live under `skills/reports/`. Evidence lives under\n`skills/reports/evidence//`.\n\nFor performance cases, inspect:\n\n- `metrics.json` for p50/p95/p99, error rate, and total duration;\n- `automation-result.json` for threshold decisions and artifacts;\n- `console.log` and `network.log` for frontend/API failures;\n- backend logs for provider, runner, WebSocket, or persistence failures.\n\nDo not call a user-path performance result a LangBot overhead regression until\nprovider/tool/network time has been separated or ruled out.\n","skills/package.json":"{\n \"private\": true,\n \"type\": \"module\",\n \"bin\": {\n \"lbs\": \"./bin/lbs\"\n },\n \"scripts\": {\n \"bootstrap\": \"node scripts/bootstrap-lbs.mjs\",\n \"prepare\": \"node scripts/bootstrap-lbs.mjs\",\n \"prevalidate\": \"node scripts/bootstrap-lbs.mjs\",\n \"preindex\": \"node scripts/bootstrap-lbs.mjs\",\n \"preindex:check\": \"node scripts/bootstrap-lbs.mjs\",\n \"pretest\": \"node scripts/bootstrap-lbs.mjs\",\n \"precheck\": \"node scripts/bootstrap-lbs.mjs\",\n \"lbs\": \"node src/lbs.ts\",\n \"test\": \"node test/lbs-cli.test.ts && python3 -m unittest discover -s test -p 'test_*.py'\",\n \"validate\": \"node src/lbs.ts validate\",\n \"index\": \"node src/lbs.ts index\",\n \"index:check\": \"node src/lbs.ts index --check\",\n \"check:syntax\": \"find src test scripts -type f \\\\( -name '*.ts' -o -name '*.mjs' \\\\) -print0 | xargs -0 -n1 node --check\",\n \"check\": \"npm run check:syntax && npm run validate && npm test\"\n },\n \"engines\": {\n \"node\": \">=22.6\"\n },\n \"devDependencies\": {\n \"playwright\": \"^1.60.0\"\n }\n}\n","skills/qa-agent-docs/qa-agent/00-technology-options.md":"# LangBot Agent Testing 技术选型\n\n## 状态\n\n这是技术选型背景文档，不是当前路线图。当前黑盒 E2E QA 的实施顺序见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n## 目标\n\n`langbot-skills` 的目标不是替代测试框架，而是沉淀 agent 可复用的测试资产，让开发者 clone 仓库后，可以让 Codex、Claude Code、Computer Use 或 Playwright MCP 复用已有路径完成 LangBot 功能验证。\n\n核心原则：\n\n- Skill 负责路由和少量规则。\n- Reference 负责可读流程和背景知识。\n- Case 负责结构化测试路径。\n- Troubleshooting 负责结构化故障资产。\n- `lbs` 负责结构校验、索引、资产创建和未来的运行/报告能力。\n- UI/browser 是产品 QA 的主路径；API/curl 只用于诊断。\n\n## 浏览器控制层\n\n不同开发者可用的浏览器控制能力不同，所以浏览器层必须可替换。\n\n| 方案 | 适用场景 | 优点 | 代价 |\n|---|---|---|---|\n| Codex / Claude Computer Use | agent 可以直接控制可见浏览器 | 登录和交互路径最自然，通常不需要额外 MCP 浏览器桥接 | 依赖具体 agent 工具能力 |\n| Playwright MCP | 没有 Computer Use，但有 MCP 浏览器工具 | 稳定、可脚本化、适合回归路径 | OAuth 登录通常需要额外 visible profile |\n| 直接 Playwright 脚本 | 测试路径非常稳定，适合 CI | 可重复性强 | 需要维护脚本和 selector |\n| 商业 AI QA 平台 | 团队希望外包测试运行平台 | 报告和 PR 集成完整 | 成本和平台绑定 |\n\n## 当前推荐\n\n先采用分层降级：\n\n```text\n有 Computer Use？\n 是 -> 使用 Computer Use 控制浏览器\n 否 -> 使用 Playwright MCP\n\n需要 GitHub OAuth？\n 是 -> 使用持久浏览器 profile，让用户手动完成登录\n 否 -> 直接使用已有登录态或测试账号状态\n```\n\n具体选择逻辑沉淀在：\n\n```text\nskills/langbot-env-setup/references/browser-access-selection.md\n```\n\n测试原则固定在：\n\n```text\ndocs/qa-agent/03-agent-browser-qa-principles.md\n```\n\n## 环境变量层\n\n测试文档不应写死端口。共享默认值放在：\n\n```text\nskills/.env\n```\n\n关键变量：\n\n```text\nLANGBOT_FRONTEND_URL\nLANGBOT_BACKEND_URL\nLANGBOT_DEV_FRONTEND_URL\nLANGBOT_REPO\nLANGBOT_WEB_REPO\nLANGBOT_BROWSER_PROFILE\n```\n\nAgent 执行测试前应先读取 `skills/.env`，再用用户提供的当前环境或已启动服务覆盖默认值。\n\n## 测试资产层\n\n测试资产分两类：\n\n```text\nskills//\n references/ # Markdown 流程说明\n cases/ # 结构化测试用例\n troubleshooting/ # 结构化故障记录\n```\n\n当前已实现：\n\n- `SKILL.md` 路由\n- `references/*.md`\n- `lbs case new/list/show`\n- `lbs trouble show/search`\n- `lbs test plan`\n- `lbs test report`\n- `lbs list / validate / index`\n\n下一步重点：\n\n- 日志守卫规则补充\n- 报告产物管理\n\n## 关键判断\n\n不要强制所有内容只能通过 CLI 修改。更好的模式是：\n\n- 新增 case/troubleshooting：优先使用 `lbs`\n- 大段流程说明：允许直接编辑 Markdown\n- 结构性变更后：必须运行 `lbs validate`\n- 任何生成索引的变更后：运行 `lbs index`\n\n这样既能沉淀结构化资产，又不会在 schema 未稳定时拖慢迭代。\n","skills/qa-agent-docs/qa-agent/01-qa-agent-harness-plan.md":"# LangBot Skills 测试资产库规划\n\n## 状态\n\n这是早期测试资产库规划文档，保留用于解释 `langbot-skills` 的分层来源。\n\n当前路线已经收敛为黑盒 E2E QA：开发者用 agent 通过浏览器测试 LangBot，\n稳定路径沉淀为 case，失败知识沉淀为 troubleshooting。`lbs test report` 和\n日志守卫已有 MVP，后续重点是报告证据、case 元数据和少量稳定路径自动化。当前优先级见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n本文中关于 `case list/show`、`trouble show/search`、`test plan` 的“计划实现”\n内容已经部分过时，因为这些能力已经落地。\n\n## 目标\n\n让开发者 clone `langbot-skills` 后，可以把测试意图交给 agent，由 agent 复用已有环境配置、测试路径和故障知识完成 LangBot 功能验证。\n\n典型场景：\n\n- 冒烟测试：验证 pipeline Debug Chat、provider、常见页面是否正常。\n- Provider 测试：添加 DeepSeek/OpenAI/Claude 等供应商并验证模型可用。\n- 新 feature 测试：探索新 UI 路径，并在稳定后沉淀成 case/reference。\n- 回归测试：复用旧路径，避免每个窗口重新探索登录、模型配置、pipeline 调试。\n- 故障沉淀：把 runtime 超时、代理不一致、WebSocket 问题记录为可搜索资产。\n\n核心方向见 `03-agent-browser-qa-principles.md`：agent 必须以浏览器/UI 为主路径，API/curl 只能作为诊断手段。\n\n## 当前仓库结构\n\n```text\nskills/\n .env # 共享默认变量\n langbot-env-setup/ # 环境准备、浏览器控制路径、代理、登录态\n langbot-testing/ # WebUI / provider / pipeline 测试入口\n langbot-plugin-dev/ # 插件开发测试\n langbot-eba-adapter-dev/ # 平台适配器开发测试\nsrc/\n lbs.ts # CLI 源码\nbin/\n lbs # CLI 入口\ndocs/\n qa-agent/ # 规划文档，历史目录名保留\n```\n\n## 设计分层\n\n### 1. Skill 层\n\n`SKILL.md` 只做触发和路由，不承载大段流程。\n\n例子：\n\n```text\nlangbot-env-setup -> 选择 Computer Use / Playwright MCP / OAuth profile / proxy\nlangbot-testing -> 选择 WebUI / pipeline / provider / troubleshooting\n```\n\n### 2. Reference 层\n\nMarkdown 记录人和 agent 都能读的流程说明。\n\n适合内容：\n\n- 如何选择浏览器控制方式\n- 如何启动/检查服务\n- 如何执行 pipeline Debug Chat\n- 如何处理 OAuth 登录态\n\n### 3. Case 层\n\n使用 YAML 记录可重复测试路径。\n\n建议结构：\n\n```text\nskills/langbot-testing/cases/\n pipeline-debug-chat.yaml\n provider-deepseek.yaml\n```\n\n建议格式：\n\n```yaml\nid: pipeline-debug-chat\ntitle: Pipeline Debug Chat returns a bot response\nmode: agent-browser\narea: pipeline\ntype: smoke\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nsteps:\n - Open LANGBOT_FRONTEND_URL\n - Navigate to Pipelines\n - Open target pipeline\n - Select Debug Chat\n - Send deterministic prompt\nchecks:\n - \"UI: User message appears\"\n - \"UI: Bot message appears\"\n - \"Console: No unexpected frontend errors\"\n - \"Logs: Backend log includes Conversation(0) Streaming completed\"\ndiagnostics:\n - \"Use API/curl only after the UI path is attempted, to distinguish frontend display failure from backend/runtime failure.\"\ntroubleshooting:\n - plugin-runtime-timeout\n - proxy-env-mismatch\n```\n\n### 4. Troubleshooting 层\n\n故障资产会逐渐变大，适合结构化记录。\n\n历史 Markdown 入口保留在：\n\n```text\nskills/langbot-testing/references/troubleshooting.md\n```\n\n当前 canonical 结构化故障资产在：\n\n```text\nskills/langbot-testing/troubleshooting/\n plugin-runtime-timeout.yaml\n proxy-env-mismatch.yaml\n```\n\n### 5. CLI 层\n\n`lbs` 是统一入口，不再引入独立 `qa` 命令。\n\n已实现或当前可用：\n\n```bash\nbin/lbs list\nbin/lbs validate\nbin/lbs index\nbin/lbs new-skill \nbin/lbs new-ref \nbin/lbs case new pipeline-debug-chat --title \"Pipeline Debug Chat\"\nbin/lbs case list\nbin/lbs case show pipeline-debug-chat\nbin/lbs trouble list \nbin/lbs trouble show plugin-runtime-timeout\nbin/lbs trouble search runtime\nbin/lbs trouble add --title ... --symptom ... --cause ... --fix ...\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log\n```\n\n## 测试库位置\n\n不要使用隐藏 `.qa/` 作为主测试库。测试资产应该和 skill 放在一起，便于触发和维护：\n\n```text\nskills/langbot-testing/\n references/\n cases/\n troubleshooting/\n reports/ # 可选，本地运行产物可按需忽略或输出到外部目录\n```\n\n如果未来需要项目本地测试库，可以允许 `lbs` 支持 `--workspace` 或项目根目录配置，但 canonical 资产仍保存在 `langbot-skills`。\n\n## 阶段规划\n\n### 阶段一：环境和测试路径沉淀\n\n状态：基本完成，持续维护。\n\n- `skills/.env` 管共享默认变量。\n- `langbot-env-setup` 拆出 Computer Use、Playwright MCP、OAuth profile、proxy、service startup。\n- `langbot-testing` 记录 WebUI、pipeline、provider 测试路径。\n- `lbs validate/index` 维护结构。\n\n完成标准：\n\n- agent 可以从 `skills/.env` 和 references 中找到当前测试入口。\n- pipeline Debug Chat 这类路径不再需要从头探索。\n\n### 阶段二：结构化 case/troubleshooting\n\n状态：主体已完成，继续补齐元数据和资产质量。\n\n目标：\n\n- `lbs case new/list/show`\n- `lbs trouble show/search`\n- case id 去重、字段校验、索引生成\n\n完成标准：\n\n- 冒烟测试路径可以用结构化 case 表示。\n- 下一个 agent 窗口可以直接读取 case 执行。\n\n### 阶段三：计划和报告\n\n状态：已有 MVP，继续完善。\n\n目标：\n\n- `lbs test plan `\n- agent 按 plan 使用浏览器执行 UI QA\n- `lbs test report`\n- 日志守卫集成\n- 报告产物和 evidence 约定\n\n完成标准：\n\n- agent 可以按 case plan 执行浏览器测试。\n- 结果报告包含 UI 结果、后端日志、console 错误和 troubleshooting 建议。\n\n## 执行规则\n\n- agent 可以直接编辑 Markdown reference。\n- 新增结构化 case/troubleshooting 时，优先使用 `lbs`。\n- 每次结构变更后运行 `bin/lbs validate`。\n- 每次索引相关变更后运行 `bin/lbs index`。\n- 测试文档不写死端口，使用 `skills/.env` 中的 URL 变量。\n- 测试 case 的 `mode` 固定为 `agent-browser`。\n- API/curl 只能写入 `diagnostics`，不能替代 UI 步骤和 UI 检查。\n","skills/qa-agent-docs/qa-agent/02-log-guard-plan.md":"# 日志守卫规划\n\n## 状态\n\n这是当前活跃设计，已有第一版文件扫描 MVP。实现边界需要和黑盒 E2E 路线保持一致：\n\n- 日志守卫服务于 `lbs test report`。\n- 它不替代浏览器/UI 判断。\n- 它不发展成独立后端 API 测试框架。\n- 第一版默认扫描 `LANGBOT_REPO/data/logs/` 下最新的 `langbot-*.log`，也可扫描 agent\n 显式提供的 backend/frontend/console 日志文件。\n\n当前总体路线见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n## 目标\n\n日志守卫是 `lbs test report` 的一部分，用来在 agent 执行测试期间捕获 UI 断言之外的运行时问题。\n\n当前命令方向已收敛为 `lbs test plan` / `lbs test report`。日志守卫服务于 agent-browser QA，不是独立的后端 API 测试入口。\n\nLangBot 是异步且集成度高的系统，有些问题不会直接表现为页面失败：\n\n- 后台任务异常\n- 未等待的协程\n- Provider 流式调用失败\n- 插件 runtime 超时\n- 平台发送失败\n- 数据库连接问题\n- 敏感信息泄露\n\n日志守卫负责把这些信号结构化地放进测试报告，并关联到 troubleshooting 资产。\n\n## 输入\n\n日志守卫应从环境和运行上下文读取配置：\n\n- `skills/.env` 中的 `LANGBOT_BACKEND_URL`\n- `skills/.env` 中的 `LANGBOT_REPO`，用于自动发现 LangBot 后端日志\n- `lbs test plan` / report 记录的 case id\n- LangBot 后端进程输出\n- 前端 dev server 输出\n- 浏览器 console/network 错误\n- case 声明的 success/failure patterns 和 expected failures\n\n## MVP 范围\n\n- 读取一个或多个日志流或日志文件。\n- 检测错误模式。\n- 支持按 case id 或 pattern 白名单。\n- 输出 JSON/Markdown 摘要。\n- 发现非预期错误时让测试报告标记失败；未来如果有自动执行器，再返回非零退出码。\n\n## 错误分类\n\n### 永远非预期\n\n除非 case 明确声明，否则应失败：\n\n- `Traceback`\n- `Task exception was never retrieved`\n- `RuntimeWarning: coroutine .* was never awaited`\n- `Unclosed client session`\n- `Unclosed connector`\n- `KeyError`\n- `TypeError`\n- `AttributeError`\n- 密钥、token、secret 明文泄露\n\n### Case 预期错误\n\n只有当前 case 声明时允许：\n\n- 无效 provider key\n- Provider 认证失败\n- 无效 webhook payload\n- 插件测试故意抛错\n- 超时测试\n- 限流测试\n\n### 仅警告\n\n报告但默认不失败：\n\n- 可恢复重试\n- 恢复的超时\n- 废弃配置\n- 慢请求\n- 版本检查失败\n\n## 与 Troubleshooting 集成\n\n日志守卫不只输出错误文本，还应尽量匹配已知 troubleshooting id。\n\n例子：\n\n```text\nAction list_plugins call timed out\nAction list_runners call timed out\nAction invoke_llm_stream call timed out\n```\n\n可映射到：\n\n```text\nplugin-runtime-timeout\n```\n\n```text\nuppercase proxy points to one host, lowercase proxy points to another\n```\n\n可映射到：\n\n```text\nproxy-env-mismatch\n```\n\n## 未来命令\n\n```bash\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report --output report.md\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log --console-log /path/to/console.log\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs test report pipeline-debug-chat --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\" --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --no-auto-log\n```\n\n运行报告应包含：\n\n- case id\n- URL 和环境变量摘要，不能包含 secrets\n- 浏览器可见结果\n- 后端日志摘要\n- console/network 错误\n- 匹配到的 troubleshooting id\n- 通过/失败结论\n\n## MVP 完成标准\n\n- 可以自动扫描最新 LangBot 后端日志，也可以扫描前端日志和 console 日志文件。\n- 可以用 `--since` 或 `--tail-lines` 把扫描范围限制到本次测试窗口。\n- 可以检测明显 Python/运行时错误和 secret 泄露风险。\n- 可以识别 case 声明的 success/failure patterns。\n- 可以识别 troubleshooting pattern，包括 `plugin-runtime-timeout` 和 `proxy-env-mismatch`。\n- 支持 case 级白名单。\n- 输出机器可读摘要。\n- 至少一个 `langbot-testing` case 使用它。\n\n当前 MVP 已覆盖自动发现 LangBot 后端日志、文件扫描、`--since`/`--tail-lines` 扫描窗口、\n基础错误检测、case success/failure signal、troubleshooting 匹配、secret 脱敏和 `--json`\n输出。仍待继续完善的是 live log 采集、更多规则、case 级 expected failure 的资产化和真实\nE2E report 样例。\n","skills/qa-agent-docs/qa-agent/03-agent-browser-qa-principles.md":"# Agent Browser QA Principles\n\nThis document fixes the direction of LangBot agent testing so the project does not drift into a backend API smoke-test framework.\n\n## Primary Goal\n\n`langbot-skills` should help an agent behave like a QA engineer using the product, not like a backend curl script.\n\nThe primary path is:\n\n```text\ndeveloper intent -> lbs test plan -> agent controls browser -> UI result + console + logs -> report/assets\n```\n\n## Rules\n\n1. Browser/UI interaction is the source of truth for product QA cases.\n2. A backend API or curl response is never enough to mark a UI case passed.\n3. API/curl/log checks are allowed as diagnostics after a UI path is attempted or when debugging environment readiness.\n4. A case passes only when the user-visible UI result is correct.\n5. The agent should inspect browser console/network output when available.\n6. If screenshot or vision capability is available, the agent should check for blank pages, overlap, hidden actions, broken layout, and error toasts.\n7. If no visual model is available, use DOM/accessibility snapshots and console output instead.\n8. New stable UI paths should be added as `cases/*.yaml`.\n9. New recurring failure modes should be added as `troubleshooting/*.yaml`.\n10. Secrets, tokens, API keys, and localStorage token values must never be printed.\n\n## Command Semantics\n\n`lbs` manages assets and produces plans. It does not replace the agent's browser-control ability.\n\n```bash\nbin/lbs test plan pipeline-debug-chat\n```\n\nThis command outputs:\n\n- environment variables to use\n- required skills\n- browser steps\n- UI/console/visual/log checks\n- diagnostic options\n- related troubleshooting patterns\n- report template\n\nThe active agent then executes the plan with Computer Use, Playwright MCP, or another available browser-control tool.\n\n## Diagnostics\n\nDiagnostics can include:\n\n- `bin/lbs env doctor`\n- browser console/network inspection\n- backend logs\n- targeted API/curl checks\n\nDiagnostics answer \"where did it fail?\" They do not replace \"did the user-visible UI work?\"\n","skills/qa-agent-docs/qa-agent/04-black-box-e2e-roadmap.md":"# 黑盒 E2E QA 路线图\n\n## 定位\n\nLangBot 有大量外部依赖：模型供应商、plugin runtime、浏览器登录态、\nmarketplace/network、RAG engine、sandbox backend、平台适配器等。单测仍然有价值，\n但这个 QA 方向当前不优先解决 LangBot core 的单测覆盖率问题，因为重 mock 往往不能\n真实代表产品路径。\n\n`langbot-skills` 当前目标是让黑盒 E2E 测试变得可执行、可沉淀、可复用：\n\n```text\n开发者测试意图\n-> 复用或新增 case\n-> agent 通过浏览器执行\n-> UI + console + network + log 证据\n-> report\n-> 反哺 case / troubleshooting\n```\n\n这是面向开发者的 QA 资产库。开发者可以让 agent 测一个 feature；如果路径稳定，\n就把路径正规化为 case，让下一个开发者或 QA agent 继续复用。\n\n## 非目标\n\n- 这一阶段不优先建设 LangBot core 单测覆盖率。\n- 不把 API/curl 作为 WebUI 行为的通过标准。\n- 不要求每个 case 都能进 CI。\n- 不在 report 和日志守卫有用之前急着做完整 browser runner。\n- 不把外部 provider、OAuth、marketplace 抖动直接判成产品失败，除非证据明确。\n\n## 当前状态\n\n仓库已经具备第一层基础设施：\n\n- `skills/.env` 和 `skills/.env.local` 管理测试环境；\n- `langbot-env-setup`、`langbot-testing`、`langbot-plugin-dev` 等 skill；\n- `skills/langbot-testing/cases` 下的结构化 case；\n- `skills/langbot-testing/troubleshooting` 下的结构化故障资产；\n- RAG、多模态、plugin、MCP 等 fixture；\n- `bin/lbs validate`、`bin/lbs index`、`bin/lbs case`、`bin/lbs trouble`、\n `bin/lbs test plan`、`bin/lbs test start`、`bin/lbs test report`。\n\n所以当前已经不是“先把路径写进 Markdown”的阶段，而是进入“让每次运行有证据、\n有报告、能沉淀”的阶段。\n\n## 测试模型\n\nUI case 只有在用户可见行为正确时才能通过。辅助证据必须解释同一次运行。\n\n通过一个 UI case 的最低证据：\n\n- 用户可见的成功信号，例如 bot 回复、provider 保存成功、文件上传完成、plugin 页面渲染；\n- 没有意外 browser console error；\n- 相关时间窗口内没有意外后端/runtime 错误；\n- 有截图、DOM snapshot 或同等视觉/结构证据，如果当前 agent 能获取；\n- API/curl 只在解释同一条 UI 路径时作为诊断证据。\n\n失败报告需要保留足够信息，让开发者能复现或分流：\n\n- case id 和实际测试 URL；\n- 使用的 browser path；\n- 最后可见 UI 状态；\n- console/network 症状；\n- 相关后端/前端日志；\n- 匹配到的 troubleshooting id；\n- 这是产品失败、环境问题、外部依赖抖动，还是证据不足。\n\n## 结果词汇\n\n统一使用这些结果：\n\n- `pass`：UI 行为正确，辅助证据干净。\n- `fail`：UI 行为错误，或同一次运行的 console/log 出现意外产品错误。\n- `blocked`：缺登录、缺 provider credentials、服务未启动等原因导致目标路径没有跑起来。\n- `env_issue`：失败在目标行为之外，例如 proxy、OAuth、provider quota、marketplace outage、\n 本地服务启动问题。\n- `flaky`：同一环境下结果不稳定，进入门禁前需要先稳定。\n\n做 merge/release 判断时，`env_issue` 和 `blocked` 不能算产品通过。\n\n## 路线图\n\n### Phase 0：对齐文档\n\n目标：明确当前黑盒 E2E 方向。\n\n交付物：\n\n- `docs/qa-agent/README.md` 文档状态导航；\n- 本路线图；\n- 给旧规划文档加状态说明。\n\n完成标准：\n\n- 新贡献者不用通读所有旧文档，也能知道当前重点。\n\n### Phase 1：Test Report MVP\n\n状态：已有第一版。\n\n目标：让每次 agent browser 测试都有一致报告格式，即使 browser 执行还没自动化。\n\n建议命令：\n\n```bash\nbin/lbs test start \nbin/lbs test report --output reports/-.md\n```\n\nMVP 行为：\n\n- 读取 case 和关联 troubleshooting；\n- 生成 Markdown report 模板；\n- 生成 run handoff，固定本次测试的 start timestamp 和推荐 report command；\n- 写入脱敏后的环境摘要；\n- 提供 `pass/fail/blocked/env_issue/flaky` 结果选项；\n- 包含 UI result、console errors、network symptoms、logs、screenshots、\n diagnostics、matched troubleshooting、assets to update 等 section；\n- 支持 `--json`，输出机器可读报告。\n\n第一版已经是 report generator，不急着做自动判定。先把 evidence 收集格式统一起来，\n再做自动化更稳。\n\n完成标准：\n\n- agent 可以先跑 `lbs test start `，用它给出的时间窗口执行浏览器路径，\n 然后按固定格式填写 report，不需要每次重新发明报告结构。\n\n### Phase 2：日志守卫 MVP\n\n状态：已有第一版文件扫描。\n\n目标：捕获 UI 不一定明显展示的 runtime 问题。\n\n日志守卫应集成进 `lbs test report`，不要发展成独立后端 API 测试框架。\n\n建议命令形态：\n\n```bash\nbin/lbs test report \\\n --backend-log /path/to/backend.log \\\n --frontend-log /path/to/frontend.log \\\n --console-log /path/to/console.log \\\n --evidence-dir reports/evidence/ \\\n --since \"2026-05-21T10:30:00+08:00\" \\\n --tail-lines 2000 \\\n --output reports/-.md\n```\n\nMVP 行为：\n\n- 默认从 `LANGBOT_REPO/data/logs/` 扫描最新 `langbot-*.log`；\n- 支持 agent 显式提供 backend、frontend、console 日志文件；\n- 支持读取 evidence 目录下的 `automation-result.json`，把浏览器自动化脚本结论纳入报告；\n- 支持 `lbs test result` 为人工/agent browser 运行写入标准 `result.json`，供 suite 聚合；\n- 支持 `--since` 和 `--tail-lines`，避免历史日志污染本次报告；\n- 检测默认非预期模式，例如 `Traceback`、未 await coroutine、unclosed client/connector、\n `KeyError`、`TypeError`、`AttributeError`、明显 secret 泄露；\n- 匹配 case 声明的 `success_patterns` 和 `failure_patterns`；\n- 匹配已知 troubleshooting，先支持 `plugin-runtime-timeout` 和 `proxy-env-mismatch`；\n- 只有 case 明确声明时，才允许 expected failure；\n- 将发现分类为 fail、warning、matched troubleshooting、ignored expected issue；\n- 永远不打印 secret 值。\n\n完成标准：\n\n- 至少 `pipeline-debug-chat` 能生成包含日志摘要和 troubleshooting 匹配结果的 report。\n\n### Phase 3：Case 元数据加固\n\n状态：已有第一版。\n\n目标：让 case 更容易选择、执行和晋级。\n\n字段逐步补充，保持向后兼容：\n\n```yaml\npriority: p0 | p1 | p2\nrisk: low | medium | high\nci_eligible: false\npreconditions:\n - \"Authenticated browser profile is available.\"\nsetup:\n - \"Start LangBot backend and frontend.\"\ncleanup:\n - \"Remove temporary provider, plugin, or knowledge base if created.\"\nexpected_failures: []\nsuccess_patterns:\n - \"Conversation(0) Streaming completed\"\nfailure_patterns:\n - \"Action invoke_llm_stream call timed out\"\nevidence:\n required:\n - ui\n - console\n - backend_log\n```\n\n当前实现采用扁平字段 `evidence_required`，避免轻量 YAML 解析器在 case 文件里承载嵌套结构。\n`bin/lbs validate` 会校验 `priority`、`risk`、`ci_eligible`、`evidence_required`、\n`automation` 脚本路径、case 关联 skill 和 troubleshooting 交叉引用。`bin/lbs case list`\n支持 `--json`、`--type`、`--area`、`--tag`、`--priority`、`--risk`、`--automation`、`--ci`\n、`--ready` 和 `--machine-ready` 过滤，方便 agent 快速选择测试集。\n`env_any` 和 `automation_env_any` 用于表达 URL-or-name 这类 one-of 输入，避免把可替代变量误判为全部必填。\n\n当前也有 `skills//suites/*.yaml` 和 `bin/lbs suite plan `，用于组织常跑测试集，\n例如 `core-smoke`、`local-agent-gate` 和\n`agent-runner-release-gate`。发布门禁使用 `agent-runner-release-preflight`\n先分类配置 blockers 和 runtime env issues，再运行较重的浏览器 Debug Chat case。\n依赖 fixture 的 case 可以在浏览器执行前先跑 `bin/lbs fixture check`，检查\n`fixtures/fixtures.json` 登记的 deterministic 文件、plugin 包和本地测试 server 是否存在。\n`bin/lbs suite start ` 会生成 suite run id、suite evidence root、per-case evidence 目录、\n`suite-start.json`/`suite-start.md` handoff 文件和 per-case evidence 命令；\n浏览器自动化脚本会写入 `automation-result.json`，供 `bin/lbs test report` 展示原始自动化结论；\n`bin/lbs test result ` 会在人工/agent browser case 完成后写入最终 `result.json`；\n`bin/lbs suite report --evidence-dir ` 会聚合各 case 的 `result.json`，并且\n不会把缺少 required evidence 的 `pass` 当作 suite 通过。\nRunner 专用 Debug Chat case 通过 `automation_pipeline_url_env` 和\n`automation_pipeline_name_env` 绑定专用 pipeline 变量，避免 local-agent、Codex 或\nClaude Code case 误用通用 `LANGBOT_PIPELINE_URL` 后产生假阳性。\nDebug Chat case 还可以通过 `automation_stream_output` 固定流式或非流式发送路径。\n多模态 Debug Chat case 可以通过 `automation_image_base64_fixture` 复用 deterministic 图片 fixture。\n`test plan` 和 `suite plan` 会输出 readiness，让 agent 在执行浏览器前就看到缺失的 env、\n自动化变量、fixture，以及需要人工确认的 `manual_check` 前置条件。\n\n完成标准：\n\n- `lbs case list` 或后续 filter 能回答“smoke 跑哪些”、“哪些适合 CI”、\n “哪些需要真实 provider credentials”。\n\n### Phase 4：开发者沉淀流程\n\n目标：开发者让 agent 测新 feature 后，稳定路径不会丢在聊天记录里。\n\n流程：\n\n1. 开发者要求 agent 通过浏览器测试某个 feature。\n2. agent 先按 UI 主路径探索。\n3. agent 用 `lbs test start` 固定运行窗口，再用 `lbs test report` 写报告。\n4. 如果路径稳定，agent 新增或更新 case。\n5. 如果出现可复用故障，agent 新增或更新 troubleshooting。\n6. agent 跑 `bin/lbs validate` 和 `bin/lbs index`。\n\n完成标准：\n\n- feature QA 的结果能进入资产库，而不是只留在一次对话里。\n\n### Phase 5：选择性浏览器自动化\n\n状态：已有第一版 `test run` 入口和两个 Playwright 脚本。\n\n目标：只自动化少量稳定、值得重复跑的黑盒路径。\n\n建议顺序：\n\n1. `webui-login-state`\n2. `pipeline-debug-chat`\n3. `local-agent-basic-debug-chat`\n4. `local-agent-rag-debug-chat`\n5. 一个基于 deterministic fixture 的 plugin 或 MCP smoke path\n\n执行策略：\n\n- 继续把 Computer Use 或 Playwright MCP 作为默认交互路径；\n- 只给稳定、确定性的路径补直接 Playwright script；\n- 保存 screenshots、console logs、trace/video；\n- flaky 或强依赖真实 credentials 的 provider case 暂时不要进 CI。\n\n当前已经绑定：\n\n- `webui-login-state` -> `scripts/e2e/webui-login-state.mjs`\n- `pipeline-debug-chat` -> `scripts/e2e/pipeline-debug-chat.mjs`\n\n第一版自动化先产出 `reports/evidence//` 下的 console、network、screenshot 和\nresult JSON。真实执行后仍要用 `lbs test report --since ... --console-log ...` 做日志守卫和\n最终报告。开发期间可以先用 `bin/lbs test run --dry-run` 检查命令和 evidence 路径。\nDebug Chat 类脚本应复用 `scripts/e2e/lib/debug-chat.mjs`，避免重复实现 visible response leaf\n判断和已知失败信号分类。\n\n完成标准：\n\n- 小规模 smoke subset 可以不靠人工决定每一步点击；更大的资产库仍然服务于人工/agent\n 驱动的探索式 E2E。\n\n## 下一批动工切片\n\n在做 browser runner 之前，继续做这些：\n\n1. 等 LangBot 当前开发状态稳定后，用一次真实 `pipeline-debug-chat` 跑通\n `test start -> test run -> test report -> test result -> suite report`，产出 sample report。\n2. 只给 smoke/local-agent 首批 case 补必要元数据。\n3. 继续补日志守卫规则，尤其是 WebSocket、plugin runtime、provider streaming、前端\n chunk/rendering failure。\n4. 约定 report 产物目录、截图和 console/network 导出的命名方式。\n5. 再评估是否开始给 `webui-login-state` 和 `pipeline-debug-chat` 做直接 Playwright\n 自动化。\n\n这样 infra 会立刻有用，同时保留后续自动化 browser execution 的空间。\n","skills/qa-agent-docs/qa-agent/README.md":"# LangBot QA Agent 文档导航\n\n这个目录记录 `langbot-skills` 当前的 QA 方向和后续建设顺序。\n\n## 当前判断\n\n当前重点是 LangBot 的黑盒 E2E QA，不是 LangBot core 的单测覆盖率建设。\n\n`langbot-skills` 要帮助开发者和 QA agent 做接近人工测试的 WebUI 验证：\n\n- 打开真实 LangBot WebUI；\n- 按用户路径点击和输入；\n- 检查用户可见的 UI 结果；\n- 查看 console、network、截图、后端和前端日志；\n- 输出可复用的测试报告；\n- 把稳定 feature 路径沉淀为 case；\n- 把重复故障沉淀为 troubleshooting。\n\nAPI 和 curl 只做诊断。它们可以解释失败原因，但不能让一个 UI case 通过。\n\n## 文档状态\n\n| 文档 | 状态 | 用途 |\n| --- | --- | --- |\n| `04-black-box-e2e-roadmap.md` | 当前主路线图 | 决定下一步建设什么。 |\n| `03-agent-browser-qa-principles.md` | 当前原则文档 | 定义 browser-first QA 的通过标准。 |\n| `02-log-guard-plan.md` | 当前活跃设计 | 设计 `lbs test report` 里的日志守卫。 |\n| `../user-guide.md` | 当前使用手册 | 开发者日常使用。 |\n| `00-technology-options.md` | 背景文档 | 选择 Computer Use、Playwright MCP 或未来直接 Playwright。 |\n| `01-qa-agent-harness-plan.md` | 历史规划，部分过时 | 解释最初分层和目录设计；使用前先看状态说明。 |\n\n## 已过时的点\n\n`01-qa-agent-harness-plan.md` 还保留早期规划状态。现在结构化 cases、\n结构化 troubleshooting、`validate`、`index`、`lbs test plan` 都已经落地。\n\n已经补上第一版 `lbs test start`、`lbs test run`、`lbs test report` 和日志守卫文件扫描。\n`webui-login-state`、`pipeline-debug-chat` 已经绑定直接 Playwright 自动化脚本。后续重点是：\n\n- 报告 evidence 字段继续打磨；\n- case success/failure signal 和日志守卫规则继续补充；\n- 报告产物和 evidence 约定；\n- 等 LangBot 当前开发状态稳定后跑真实 sample report。\n\n不要再把旧阶段列表当成当前 source of truth。后续排序以\n`04-black-box-e2e-roadmap.md` 为准。\n","skills/qa-agent-docs/user-guide.md":"# LangBot Skills 用户使用手册\n\n## 这个仓库解决什么\n\n`langbot-skills` 是给 agent 使用的 LangBot 测试资产库。开发者 clone 后，可以让 Codex、Claude Code、Computer Use 或 Playwright MCP 复用已有环境配置、测试路径和故障知识，像 QA 一样操作 LangBot WebUI。\n\n核心目标：\n\n- 不让下一个 agent 窗口从头探索登录、模型配置、pipeline 调试。\n- 把稳定 UI 测试路径沉淀为 case。\n- 把常见故障沉淀为 troubleshooting。\n- 让 agent 优先通过浏览器点击验证产品行为。\n- API/curl/log 只作为诊断手段，不作为 UI case 通过标准。\n\n## 快速开始\n\n1. Clone 仓库。\n\n2. 检查本地默认变量：\n\n```bash\nbin/lbs env show\n```\n\n默认变量在：\n\n```text\nskills/.env\n```\n\n本机专用覆盖写到：\n\n```text\nskills/.env.local\n```\n\n它会覆盖 `skills/.env` 中的同名变量，并且不应该提交。\n`skills/.env` 是共享默认值，不应写入本机绝对路径、浏览器 profile、provider key 或其他凭据。\n新机器建议从模板开始：\n\n```bash\ncp skills/.env.example skills/.env.local\n```\n\n常用变量：\n\n```text\nLANGBOT_FRONTEND_URL\nLANGBOT_BACKEND_URL\nLANGBOT_DEV_FRONTEND_URL\nLANGBOT_REPO\nLANGBOT_WEB_REPO\nLANGBOT_BROWSER_PROFILE\n```\n\n3. 检查环境是否就绪：\n\n```bash\nbin/lbs env doctor\nbin/lbs fixture check\n```\n\n`env doctor` 会检查 URL、路径、代理变量等。代理变量是可选项；只有大小写代理变量互相冲突时才会报错。失败不一定代表仓库坏了，通常说明本地 LangBot 没启动、代理不一致或浏览器 profile 不存在。\n`fixture check` 会检查仓库内测试 fixture 是否存在，例如 MCP stdio server、RAG 文档、多模态图片、qa-plugin-smoke 包和 QA Runner 包。它也会校验 `.lbpkg` 是 zip 包，并检查 QA Runner fixture 的入口文件未漂移。\n\n4. 查看已有测试 case：\n\n```bash\nbin/lbs case list\nbin/lbs case list --json --priority p0 --automation\nbin/lbs case list --ready\nbin/lbs case list --machine-ready\nbin/lbs suite list\nbin/lbs suite plan core-smoke\nbin/lbs suite plan agent-runner-release-gate\nbin/lbs suite start core-smoke\nbin/lbs suite start core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\n```\n\n`case list` 支持按 `--type`、`--area`、`--tag`、`--priority`、`--risk`、`--automation`\n、`--ci`、`--ready` 和 `--machine-ready` 过滤。`--ready` 只显示没有缺机器输入且没有人工前置条件的 case；\n`--machine-ready` 过滤掉缺机器输入的 case，但保留 `manual-check`，表示执行前还要确认前置条件。需要交给 agent 自动选择测试集时，优先使用 `--json`，\n其中包含 `priority`、`risk`、`ci_eligible`、`automation`、`evidence_required` 以及\nenv/automation/fixture/manual readiness。\nCase metadata 中的 `env` 和 `automation_env` 表示全部必填；URL 或 name 这类二选一输入会放在\n`env_any` 或 `automation_env_any`，readiness 只要求组合里至少一个变量有值。\n\n如果要跑一组已沉淀的测试路径，优先使用 suite。Suite 位于 `skills//suites/*.yaml`，\n只负责组织 case，不改变 UI/browser 作为通过标准的原则。\n`suite plan` 会聚合 readiness：缺环境变量、缺自动化变量、缺 fixture 或需要\n`manual_check` 时，会在执行前标出受影响的 case。`manual_check` 不是产品通过，\n它表示机器配置已满足但 agent 必须先确认 case 里的 `preconditions` 或 `setup`。\nRunner externalization 发布判断使用 `agent-runner-release-gate`。先跑\n`agent-runner-release-preflight`，把缺 pipeline、runner id 错误、插件未安装这类\n`blocked`，以及 provider key、Box、插件运行时这类 `env_issue` 分开，再执行较重的\n浏览器 Debug Chat case。\n\n5. 生成 agent 执行计划：\n\n```bash\nbin/lbs test plan pipeline-debug-chat\n```\n\n然后把计划交给当前 agent 执行。agent 应使用 Computer Use、Playwright MCP 或其他浏览器控制能力去操作 UI。\n`test plan` 中的 Environment、Automation Readiness、Fixture Readiness 和 Manual\nReadiness 是执行前门禁；如果 readiness 缺失，应先补配置或将本次 case 标记为\n`blocked`。如果状态是 `manual_check`，先确认 `preconditions` 和 `setup`，再开始 UI\n执行。不要把后续 curl/API 诊断当成 UI case 通过。\n\n## 推荐使用方式\n\n### 冒烟测试\n\n你可以直接对 agent 说：\n\n```text\n帮我跑一下 LangBot 新前端 smoke test。\n```\n\nagent 应该：\n\n- 读 `skills/.env`\n- 优先查看 `bin/lbs suite plan core-smoke`，或查找 `type: smoke` 的 cases\n- 生成 test plan\n- 用浏览器执行 UI 操作\n- 检查 console、截图、后端日志\n- 输出简短 QA 报告\n\n### Runner Externalization 发布门禁\n\n你可以直接对 agent 说：\n\n```text\n按 agent-runner release gate 跑完整矩阵，先做 preflight，再跑浏览器 case，并把 blocked/env_issue/fail 分开。\n```\n\nagent 应该先查看 `skills/langbot-testing/references/agent-runner-release-gate.md`，\n再执行：\n\n```bash\nbin/lbs test recommend\nbin/lbs suite plan agent-runner-release-gate\nbin/lbs test run agent-runner-release-preflight --dry-run\nbin/lbs suite start agent-runner-release-gate --run-id agent-runner-release-local --evidence-dir reports/evidence/agent-runner-release-local\n```\n\n`test recommend` 输出的 run 命令默认带 `--dry-run`；确认 readiness 和 `manual_check` 前置条件后，再去掉 `--dry-run` 执行。\n\n完成所有 case 后，用：\n\n```bash\nbin/lbs suite report agent-runner-release-gate --evidence-dir reports/evidence/agent-runner-release-local\n```\n\n没有最终 `result.json`、缺 required evidence、或把 `blocked`/`env_issue` 当成 pass，\n都不能算发布门禁通过。\n\n### 新 Feature 测试\n\n你可以说：\n\n```text\n我改了 provider 页面，帮我测一下 DeepSeek provider 添加、测试、绑定 pipeline 是否正常。\n```\n\nagent 应该：\n\n- 查找相关 case 和 reference\n- 如果没有稳定路径，先探索 UI\n- 用浏览器执行真实交互\n- 失败时用日志/API 辅助定位\n- 稳定后新增或更新 case/reference\n- 新故障沉淀为 troubleshooting\n\n### 定点排错\n\n你可以说：\n\n```text\nDebug Chat 点了没回复，帮我查是前端问题还是后端问题。\n```\n\nagent 应该：\n\n- 先通过 UI 复现\n- 看 console/network\n- 看后端日志\n- 必要时用 API/curl 做诊断\n- 匹配 troubleshooting\n- 给出修复建议或直接修复\n\n## 重要原则\n\n这些原则固定在：\n\n```text\ndocs/qa-agent/03-agent-browser-qa-principles.md\n```\n\n简化版：\n\n- UI/browser 是测试主路径。\n- API/curl/log 只做诊断。\n- 后端接口成功不等于 UI case 通过。\n- case 通过必须以用户可见 UI 结果为准。\n- 有视觉能力时应检查截图。\n- 没有视觉能力时用 DOM/accessibility snapshot 和 console。\n- 不要打印 token、API key、OAuth secret 或 localStorage token 值。\n\n## 规划文档\n\n如果要判断下一步建设什么，先看：\n\n```text\ndocs/qa-agent/README.md\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n`01-qa-agent-harness-plan.md` 是早期规划，部分内容已经被当前实现和路线图替代。\n\n## 常用命令\n\n### 环境\n\n```bash\nbin/lbs env show\nbin/lbs env show --json\nbin/lbs env doctor\nbin/lbs fixture list\nbin/lbs fixture check\nbin/lbs fixture check --json\n```\n\n`env show` 和 `env doctor` 默认会对 token、API key、password、secret 以及 URL basic auth\n做脱敏。不要把 `.env.local` 里的原始凭据复制进测试报告。\n\n### Skill 和索引\n\n```bash\nbin/lbs list\nbin/lbs validate\nbin/lbs index --check\nbin/lbs index\n```\n\n### Case\n\n```bash\nbin/lbs case list\nbin/lbs case list --type smoke\nbin/lbs case list --json --priority p1 --tag local-agent\nbin/lbs case list --ready\nbin/lbs case list --machine-ready\nbin/lbs case show pipeline-debug-chat\nbin/lbs case new my-feature --title \"My Feature Works\"\n```\n\n### Suite\n\n```bash\nbin/lbs suite list\nbin/lbs suite list --json --priority p1\nbin/lbs suite show local-agent-gate\nbin/lbs suite plan core-smoke\nbin/lbs suite plan local-agent-gate --json\nbin/lbs suite start core-smoke\nbin/lbs suite start core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\nbin/lbs suite run core-smoke --dry-run --json\nbin/lbs suite run core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\nbin/lbs suite start core-smoke --json\nbin/lbs suite report core-smoke --evidence-dir reports/evidence/\nbin/lbs suite report core-smoke --evidence-dir reports/evidence/ --json\nbin/lbs suite new my-feature-gate --title \"My Feature Gate\"\n```\n\n`suite start` 不直接控制浏览器。它生成统一 run id、suite evidence root、每个 case 的 evidence\n目录、`suite-start.json`/`suite-start.md` handoff 文件，以及每个 case 的 `test run`、`test report`\n和 `test result` 命令模板。需要固定路径时，使用 `--run-id` 和 `--evidence-dir`。\n`suite run --dry-run --json` 只预览 planned/skipped case，不创建 evidence，也不执行 automation。\n`suite run` 会顺序执行 suite 中已有 automation、机器 readiness 已满足且不需要 `manual_check` 的 case，并在最后聚合 `suite report`。\n缺 env、automation env 或 fixture 的 case 默认会跳过；确实要强制执行时，加 `--include-not-ready`。\n确认前置条件后，才用 `--include-manual-check` 执行这类 case。\n所有 case 执行完并写入最终 `result.json` 后，\n`suite report` 会读取各 case evidence 目录并汇总为 `pass`、`fail`、`blocked`、`env_issue`、\n`flaky`、`incomplete` 等状态。`pass` 必须声明已经收集 case 的全部 required evidence；\n否则 suite 会保持 `incomplete`，避免把缺证据的运行误判成通过。\n\n### Test Plan\n\n```bash\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test plan pipeline-debug-chat --json\n```\n\n### Test Start\n\n```bash\nbin/lbs test start pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat --json\n```\n\n`test start` 用于 agent 开始一次浏览器测试前记录 run id、开始时间和推荐 report 命令。\n它会把 `--since \"\"` 写进后续报告命令，减少历史日志污染本次判断。\n如果 case 绑定了自动化脚本，输出里也会包含 `test run` 命令和 evidence 目录。\n\n### Test Automation\n\n```bash\nbin/lbs test run webui-login-state --dry-run\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test run webui-login-state --run-id login-smoke --output reports/evidence/login-smoke\nbin/lbs test run pipeline-debug-chat --run-id pipeline-smoke --output reports/evidence/pipeline-smoke\n```\n\n查看当前所有带自动化脚本的 case：\n\n```bash\nbin/lbs case list --automation\nbin/lbs case list --json --automation\n```\n\n当前自动化覆盖包括登录态、通用 Pipeline Debug Chat、local-agent runner 的基础回复、\nPromptPreProcessing、RAG、plugin tool、MCP stdio tool、非流式、多模态和 RAG+多模态路径。\n不要在文档里手工维护静态 case 清单；以 `case list --automation` 和 suite 定义为准。\n\n自动化脚本位于 `scripts/e2e/`。它们会保存：\n\n- `console.log`\n- `network.log`\n- `screenshot.png`\n- `automation-result.json`\n\n新增 Debug Chat 类自动化时，优先复用 `scripts/e2e/lib/debug-chat.mjs` 中的 pipeline 打开、\nprompt 发送、visible response leaf 判断和失败信号分类，不要在新脚本里复制 DOM 扫描逻辑。\n\n脚本需要本地安装 Playwright 后才能真正执行：\n\n```bash\nnpm install\nnpx playwright install chromium\n```\n\n`pipeline-debug-chat` 通用自动化建议配置 `LANGBOT_PIPELINE_URL`。如果没有 direct URL，\n脚本会尝试通过 `LANGBOT_PIPELINE_NAME` 从 Pipelines 页面进入目标 pipeline。两者都没有时，\n该自动化会返回 `blocked`，不会伪造通过。\n\nRunner 专用 case 不应复用通用 pipeline 变量。Local Agent、Codex Runner 和\nClaude Code Runner 这类 case 会通过 `automation_pipeline_url_env` /\n`automation_pipeline_name_env` 映射到 case-specific env，例如\n`LANGBOT_LOCAL_AGENT_PIPELINE_URL`。这些 case 如果缺少专用变量，会返回 `blocked`，\n不会退回到 `LANGBOT_PIPELINE_URL`，避免跑错 pipeline 后产生假阳性。\n如果 case 声明了 `setup_automation`，只有 `bin/lbs test run ` 的真实执行路径会先运行这些 setup；\n`test plan`、`suite plan`、`case list` 和 `--dry-run` 只展示它们，不会修改本地环境。\nsetup 可以是 `case:` 或仓库内 `node:scripts/... --flag`，每一步证据会写到主 evidence 目录下的\n`setup/` 子目录。setup 失败时主 automation 不会继续执行；setup 写入 `.env.local` 后，主 automation\n会重新读取环境。用 `setup_provides_env` 声明 setup 会生成的变量，可以让 readiness 正确显示机器可准备状态。\n如果 Debug Chat case 需要固定流式/非流式路径，可以在 case 中设置\n`automation_stream_output: \"1\"` 或 `\"0\"`，脚本会在发送消息前切换 Debug Chat 的 Stream 控件。\n如果 case 需要上传图片，可以设置 `automation_image_base64_fixture` 指向仓库内的 base64 PNG fixture，\n脚本会在 evidence 目录写出临时 PNG 并通过 Debug Chat 上传控件发送。\n`bin/lbs test plan --json` 和 `bin/lbs suite plan --json`\n都会显示这些专用变量是否已配置。\n\n### Test Report 和日志守卫\n\n```bash\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report pipeline-debug-chat --output reports/pipeline-debug-chat.md\nbin/lbs test report pipeline-debug-chat \\\n --backend-log /path/to/backend.log \\\n --frontend-log /path/to/frontend.log \\\n --console-log /path/to/console.log\nbin/lbs test report pipeline-debug-chat --evidence-dir reports/evidence/pipeline-smoke\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log --json\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs test report pipeline-debug-chat --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\" --tail-lines 2000\n```\n\n`test report` 会生成报告模板，并默认从 `LANGBOT_REPO/data/logs/` 自动选择最新的\n`langbot-*.log` 作为 LangBot 后端日志扫描。也可以用 `--backend-log` 覆盖，或用\n`--no-auto-log` 只生成模板。\n\n如果提供 `--evidence-dir`，或 `--console-log` 指向 `reports/evidence//console.log`，\n报告会优先读取同目录的 `automation-result.json`，并展示自动化脚本的 `status`、`reason`、\n起止时间和目标 URL。\n\n日志守卫会扫描常见错误、secret 泄露风险、case 声明的 success/failure patterns，以及已知\ntroubleshooting pattern。它不控制浏览器，也不替代 UI 通过判断。`success_patterns`\n命中会作为通过证据写入 `success_signals`；声明了 success pattern 但本次扫描窗口没有命中，\n会给 warning；`failure_patterns` 命中会让本次日志守卫 fail。\n\n建议在执行浏览器 case 前记录当前时间，然后在报告阶段使用 `--since`。如果只想快速看\n最近日志，可以使用 `--tail-lines`。\n\n### Runtime Log Guard\n\n如果还没有进入某个具体 UI case，只是想观察 LangBot 后端日志，可以直接使用 `log`\n命令。它和 `test report` 使用同一套扫描器、secret 脱敏、troubleshooting pattern 和\ncase success/failure pattern。\n\n```bash\nbin/lbs log scan --tail-lines 300\nbin/lbs log scan --case pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs log scan --backend-log /path/to/langbot.log --json\nbin/lbs log scan --failure-pattern \"runner.tool_loop_error|Action invoke_llm_stream call timed out\" --strict\n```\n\n`log scan` 默认从 `LANGBOT_REPO/data/logs/` 自动选择最新的 `langbot-*.log`。传入\n`--case ` 后，会额外应用该 case 声明的 `success_patterns`、`failure_patterns`\n和 related troubleshooting。默认用于观察，返回码保持 0；加 `--strict` 后，`fail` 或\n`env_issue` 会返回非 0，适合脚本门禁。\n\n运行期观察可以用 `watch`：\n\n```bash\nbin/lbs log watch --case pipeline-debug-chat\nbin/lbs log watch --backend-log /path/to/langbot.log --interval-ms 500\nbin/lbs log watch --duration-ms 30000 --strict --json\n```\n\n`log watch` 默认从启动时的文件末尾开始，只观察新追加的日志；加 `--from-start` 可从文件开头扫。\n它会实时打印新命中的 findings 和 success signals。为了避免当前历史日志噪声影响观察，默认不因\n异常返回非 0；加 `--strict` 后，退出时如果看到 `fail` 或 `env_issue` 会返回非 0。\n\n给一次 QA 运行包日志窗口时，用 `guard start/stop`：\n\n```bash\nbin/lbs log guard start --run-id local-debug --case pipeline-debug-chat\n# 执行浏览器或手工测试\nbin/lbs log guard stop --run-id local-debug\n```\n\n`start` 会在 `reports/log-guards/.json` 记录起始时间、case 和当前后端日志路径；\n`stop` 会用 start/stop 时间作为扫描窗口，生成 `reports/log-guards/.md`，并默认按\nstrict guard 返回码处理。临时只想收集报告、不想让命令失败，可以加 `--no-strict`。\n\n当前 LangBot core 日志还不是完全结构化日志，runtime guard 主要依赖时间窗口和文本 pattern。\n已支持 ISO 时间戳和 LangBot 当前的 `[MM-DD HH:mm:ss.SSS]` 前缀；没有时间戳的连续行会跟随上一条\n带时间戳的日志块。如果后续 core 能提供稳定 request id、conversation id、plugin action id 或\nJSON log field，guard 可以从“时间窗口 + 文本匹配”升级为更精确的关联分析。\n\n### Test Result\n\n```bash\nbin/lbs test result pipeline-debug-chat \\\n --result pass \\\n --reason \"Debug Chat returned OK and the report log guard was clean.\" \\\n --evidence-dir reports/evidence/pipeline-smoke \\\n --started-at \"2026-05-21T10:30:00+08:00\" \\\n --evidence ui,screenshot,console,backend_log\n```\n\n`test result` 用于把一次人工/agent browser 运行的最终判断写成标准 `result.json`，\n供 `suite report` 聚合。它不会替代 UI 测试：如果写 `--result pass`，`--evidence`\n必须覆盖该 case 的 `evidence_required`，否则命令会失败。自动化脚本写\n`automation-result.json`；如果 case 还要求 backend log、API diagnostic 或 filesystem\nevidence，agent 需要在报告和诊断完成后再用 `test result` 写最终 `result.json`。\n\n### Troubleshooting\n\n```bash\nbin/lbs trouble list langbot-testing\nbin/lbs trouble show plugin-runtime-timeout\nbin/lbs trouble search runtime\nbin/lbs trouble add langbot-testing --title \"...\" --symptom \"...\" --cause \"...\" --fix \"...\"\n```\n\n## 目录说明\n\n```text\nskills/\n .env # 共享默认变量\n langbot-env-setup/ # 环境、浏览器、登录态、代理\n langbot-testing/ # WebUI / provider / pipeline 测试\nschemas/ # 结构化资产 schema\nsrc/ # lbs TypeScript 源码\nbin/ # lbs 入口\ndocs/ # 设计文档和用户手册\nAGENTS.md # agent 维护协议\n```\n\n## 添加一个新测试路径\n\n1. 先让 agent 通过浏览器探索并执行路径。\n2. 稳定后创建 case：\n\n```bash\nbin/lbs case new provider-xxx --title \"Provider XXX can be configured\" --area provider --type provider\n```\n\n3. 编辑生成的 `cases/*.yaml`，补充真实步骤、检查点和 troubleshooting。\n\n4. 校验：\n\n```bash\nbin/lbs validate\nbin/lbs index --check\nbin/lbs index\n```\n\n## 添加一个新故障\n\n```bash\nbin/lbs trouble add langbot-testing \\\n --title \"Plugin runtime actions time out\" \\\n --symptom \"Debug Chat shows Agent runner temporarily unavailable\" \\\n --cause \"Old plugin runtime survived backend restart\" \\\n --fix \"Stop old runtime processes and restart LangBot\"\n```\n\n然后编辑生成的 YAML，补充 `patterns`、`related_cases` 和验证方式。\n\n## 当前边界\n\n- `lbs test plan` 只生成测试计划，不直接控制浏览器。\n- `lbs test report` 生成报告，默认扫描最新 LangBot 后端日志；也可扫描显式提供的\n backend/frontend/console 日志文件。\n- 真正的 UI 操作由当前 agent 的浏览器能力执行。\n- `env doctor` 是 readiness check，不是产品测试。\n- `curl/API` 是诊断工具，不是主要测试路径。\n","skills/schemas/README.md":"# Schemas\n\n这个目录存放 LangBot skills 结构化资产的 JSON Schema。\n\n它们不是测试脚本，也不会执行浏览器动作。它们的作用是定义 agent 和维护者后续新增资产时应该遵守的文件结构。\n\n## 文件说明\n\n- `skills//fixtures/fixtures.json`\n 不是 JSON Schema，但由 `bin/lbs validate` 校验。\n 它登记 deterministic fixture 文件、类型和关联 case，供 `bin/lbs fixture check` 做 readiness 检查。\n\n- `case.schema.json`\n 约束 `skills//cases/*.yaml` 的格式。\n Case 描述 agent-browser 或 probe QA 路径，包括前置条件、步骤、检查点、诊断手段和关联故障。\n\n- `suite.schema.json`\n 约束 `skills//suites/*.yaml` 的格式。\n Suite 只组织 case 集合，用于 smoke、regression 或 release gate 等测试入口。\n\n- `troubleshooting.schema.json`\n 约束 `skills//troubleshooting/*.yaml` 的格式。\n Troubleshooting 条目描述症状、日志/错误模式、可能原因、修复步骤和验证信号。\n\n- `skill-index.schema.json`\n 约束生成文件 `skills.index.json` 的格式。\n 这个索引用于让 agent 快速发现已有 skills、references、cases、suites 和 troubleshooting。\n\n- `reports/evidence//result.json`\n 不是 catalog schema，而是执行期最终裁定产物，由 `bin/lbs test result` 写入。\n `suite report` 读取其中的 `status`、`reason`、起止时间和 `evidence_collected`，\n 并用 `evidence_missing` 防止缺证据的 `pass` 被当作完整通过。\n\n- `reports/evidence//automation-result.json`\n 不是 catalog schema，而是浏览器自动化脚本的原始运行结论，供 `bin/lbs test report`\n 展示和推断日志扫描窗口。\n\n## 为什么需要 schemas\n\nSchemas 是基础设施护栏：\n\n- 防止 case、suite 和 troubleshooting 随着增长变得格式混乱\n- 让 `bin/lbs validate` 能发现缺字段和错误结构\n- 为未来编辑器提示和 CI 校验留接口\n- 帮助 agent 新增资产时知道应该写哪些字段\n\n## 当前校验方式\n\n`bin/lbs validate` 做轻量、schema 对齐的校验，不引入额外依赖。它会检查必填字段、\n枚举值、boolean 字段、重复列表项、automation 脚本存在性，以及 case、suite、skill、\ntroubleshooting 之间的交叉引用。这里的 schema 仍是格式契约；如果未来引入正式 JSON\nSchema validator，应继续保持这些本地交叉引用检查。\n\nCase 里的 `env` / `automation_env` 表示所有列出的变量都需要配置。遇到二选一输入时，\n使用 `env_any` / `automation_env_any`，每一项写成 `LANGBOT_PIPELINE_URL|LANGBOT_PIPELINE_NAME`\n这类 one-of 组合，避免 agent 因为只配置了 URL 或 name 其中之一而误判未就绪。\n`setup` 和 `preconditions` 是人工确认项，会让 readiness 进入 `manual_check`；\n`setup_automation` 是 `test run` 可以自动执行的准备步骤，配合 `setup_provides_env`\n声明它会生成的机器变量。\n","skills/schemas/case.schema.json":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/case.schema.json\",\n \"title\": \"LangBot Skill Test Case\",\n \"type\": \"object\",\n \"required\": [\n \"id\",\n \"title\",\n \"mode\",\n \"area\",\n \"type\",\n \"priority\",\n \"risk\",\n \"ci_eligible\",\n \"tags\",\n \"skills\",\n \"steps\",\n \"checks\",\n \"evidence_required\"\n ],\n \"allOf\": [\n {\n \"if\": {\n \"properties\": {\n \"mode\": { \"const\": \"agent-browser\" }\n }\n },\n \"then\": {\n \"required\": [\"env\"]\n }\n }\n ],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"mode\": {\n \"type\": \"string\",\n \"enum\": [\"agent-browser\", \"probe\"]\n },\n \"area\": {\n \"type\": \"string\"\n },\n \"type\": {\n \"type\": \"string\",\n \"enum\": [\n \"smoke\",\n \"regression\",\n \"feature\",\n \"provider\",\n \"exploratory\",\n \"contract\",\n \"performance\",\n \"reliability\",\n \"chaos\",\n \"security\"\n ]\n },\n \"priority\": {\n \"type\": \"string\",\n \"enum\": [\"p0\", \"p1\", \"p2\"]\n },\n \"risk\": {\n \"type\": \"string\",\n \"enum\": [\"low\", \"medium\", \"high\"]\n },\n \"ci_eligible\": {\n \"type\": \"boolean\"\n },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"skills\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"env_any\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*(\\\\|[A-Z][A-Z0-9_]*)+$\"\n }\n },\n \"steps\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"checks\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"evidence_required\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"enum\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"backend_log\",\n \"frontend_log\",\n \"api_diagnostic\",\n \"filesystem\",\n \"metrics\",\n \"trace\",\n \"profile\",\n \"resource_log\"\n ]\n },\n \"minItems\": 1\n },\n \"preconditions\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup_automation\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^(?:case:[a-z0-9][a-z0-9_-]*|node:scripts/[A-Za-z0-9_./-]+\\\\.(?:mjs|js|ts)(?:\\\\s+--[A-Za-z0-9][A-Za-z0-9_-]*(?:=[A-Za-z0-9_./:@-]+)?)*)$\"\n }\n },\n \"setup_provides_env\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n }\n },\n \"cleanup\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"diagnostics\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation\": {\n \"type\": \"string\"\n },\n \"automation_env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation_env_any\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*(\\\\|[A-Z][A-Z0-9_]*)+$\"\n }\n },\n \"automation_prompt\": {\n \"type\": \"string\"\n },\n \"automation_prompts_json\": {\n \"type\": \"string\"\n },\n \"automation_expected_text\": {\n \"type\": \"string\"\n },\n \"automation_response_timeout_ms\": {\n \"type\": \"string\"\n },\n \"automation_stream_output\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_image_base64_fixture\": {\n \"type\": \"string\"\n },\n \"automation_runner_config_patch_json\": {\n \"type\": \"string\"\n },\n \"automation_restore_runner_config\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_expected_runner_id\": {\n \"type\": \"string\"\n },\n \"automation_reset_debug_chat\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_session_type\": {\n \"type\": \"string\",\n \"enum\": [\"person\", \"group\"]\n },\n \"automation_debug_chat_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_max_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_requests\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_concurrency\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_timeout_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_first_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_max_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_error_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_ok_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_provider_fault_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_expected_prefix\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_prompt_template\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_stream\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_reset\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_fail_on_final_mismatch\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_require_success\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_provider_model_thresholds_json\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_pipeline_name\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_model_name\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fallback_model_names\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_response_text\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_first_token_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_chunk_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_chunk_count\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_first_n\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_every_n\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_models\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fault_status\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_after_first_chunk\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_fake_provider_fail_after_first_chunk_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_after_first_chunk_mode\": {\n \"type\": \"string\",\n \"enum\": [\"disconnect\", \"error_event\"]\n },\n \"automation_fake_provider_fail_after_first_chunk_models\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_dynamic_response\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_filesystem_checks_json\": {\n \"type\": \"string\"\n },\n \"metrics_thresholds_json\": {\n \"type\": \"string\"\n },\n \"load_profile_json\": {\n \"type\": \"string\"\n },\n \"fault_model_json\": {\n \"type\": \"string\"\n },\n \"automation_pipeline_url_env\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n },\n \"automation_pipeline_name_env\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n },\n \"success_patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"failure_patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"expected_failures\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"troubleshooting\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n}\n","skills/schemas/skill-index.schema.json":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/skill-index.schema.json\",\n \"title\": \"LangBot Skills Index\",\n \"type\": \"object\",\n \"required\": [\"generated_by\", \"skills\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"generated_by\": {\n \"type\": \"string\"\n },\n \"skills\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\n \"directory\",\n \"name\",\n \"description\",\n \"references\",\n \"cases\",\n \"case_summaries\",\n \"suites\",\n \"suite_summaries\",\n \"fixtures\",\n \"troubleshooting\",\n \"troubleshooting_summaries\"\n ],\n \"additionalProperties\": false,\n \"properties\": {\n \"directory\": { \"type\": \"string\" },\n \"name\": { \"type\": \"string\" },\n \"description\": { \"type\": \"string\" },\n \"references\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"case_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"mode\", \"area\", \"type\", \"priority\", \"risk\", \"ci_eligible\", \"tags\", \"automation\", \"setup_automation\", \"setup_provides_env\", \"evidence_required\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"mode\": { \"type\": \"string\", \"enum\": [\"agent-browser\", \"probe\"] },\n \"area\": { \"type\": \"string\" },\n \"type\": { \"type\": \"string\" },\n \"priority\": { \"type\": \"string\" },\n \"risk\": { \"type\": \"string\" },\n \"ci_eligible\": { \"type\": \"boolean\" },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation\": { \"type\": \"string\" },\n \"setup_automation\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup_provides_env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"evidence_required\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"suites\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"suite_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"description\", \"type\", \"priority\", \"tags\", \"cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"description\": { \"type\": \"string\" },\n \"type\": { \"type\": \"string\" },\n \"priority\": { \"type\": \"string\" },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"fixtures\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"kind\", \"path\", \"related_cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"kind\": { \"type\": \"string\" },\n \"path\": { \"type\": \"string\" },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"troubleshooting\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"troubleshooting_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"category\", \"related_cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"category\": { \"type\": \"string\" },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n }\n }\n }\n }\n }\n}\n","skills/schemas/suite.schema.json":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/suite.schema.json\",\n \"title\": \"LangBot Skill Test Suite\",\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"description\", \"type\", \"priority\", \"tags\", \"cases\"],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"description\": {\n \"type\": \"string\"\n },\n \"type\": {\n \"type\": \"string\",\n \"enum\": [\n \"smoke\",\n \"regression\",\n \"release_gate\",\n \"exploratory\",\n \"contract\",\n \"performance\",\n \"reliability\",\n \"chaos\",\n \"security\"\n ]\n },\n \"priority\": {\n \"type\": \"string\",\n \"enum\": [\"p0\", \"p1\", \"p2\"]\n },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n }\n }\n}\n","skills/schemas/troubleshooting.schema.json":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/troubleshooting.schema.json\",\n \"title\": \"LangBot Skill Troubleshooting Entry\",\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"symptoms\", \"patterns\", \"likely_causes\", \"fix_steps\", \"verification\"],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"date\": {\n \"type\": \"string\"\n },\n \"category\": {\n \"type\": \"string\",\n \"enum\": [\"product\", \"env_issue\", \"external_dependency\", \"blocked\", \"flaky\"]\n },\n \"symptoms\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"likely_causes\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"fix_steps\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"verification\": {\n \"type\": \"string\"\n },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n}\n","skills/skills.index.json":"{\n \"generated_by\": \"lbs\",\n \"skills\": [\n {\n \"directory\": \"langbot-deploy\",\n \"name\": \"langbot-deploy\",\n \"description\": \"Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \\\"deploy langbot\\\", \\\"langbot docker\\\", \\\"langbot compose\\\", \\\"langbot kubernetes\\\", \\\"langbot config.yaml\\\", \\\"langbot box runtime\\\", \\\"langbot global api key\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-dev\",\n \"name\": \"langbot-dev\",\n \"description\": \"Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \\\"langbot backend\\\", \\\"langbot dev\\\", \\\"langbot api\\\", \\\"add langbot endpoint\\\", \\\"langbot migration\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-eba-adapter-dev\",\n \"name\": \"langbot-eba-adapter-dev\",\n \"description\": \"Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing.\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-env-setup\",\n \"name\": \"langbot-env-setup\",\n \"description\": \"Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths.\",\n \"references\": [\n \"references/browser-access-selection.md\",\n \"references/computer-use.md\",\n \"references/oauth-browser-profile.md\",\n \"references/playwright-mcp.md\",\n \"references/proxy.md\",\n \"references/service-startup.md\",\n \"references/wsl-notes.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-mcp-ops\",\n \"name\": \"langbot-mcp-ops\",\n \"description\": \"Operate a LangBot instance through its built-in MCP (Model Context Protocol) server. Use when an AI agent needs to manage LangBot — list/create/update/delete bots, agents, pipelines, models, knowledge bases, MCP servers, and skills — over MCP instead of raw HTTP. Covers the /mcp endpoint, API-key auth (web-UI lbk_ keys and the config.yaml global key), the tool surface, and client configuration. Triggers on \\\"langbot mcp\\\", \\\"manage langbot via mcp\\\", \\\"langbot /mcp\\\", \\\"langbot mcp server\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-plugin-dev\",\n \"name\": \"langbot-plugin-dev\",\n \"description\": \"Develop, debug, and test LangBot plugins. Use when creating new LangBot plugins, fixing plugin bugs, setting up a LangBot test environment, or testing plugins via WebSocket. Covers plugin component architecture (EventListener, Command, Tool), the plugin SDK API (invoke_llm, get_llm_models, send_message, plugin storage), common pitfalls, and automated WebSocket-based testing. Triggers on \\\"langbot plugin\\\", \\\"lbp\\\", \\\"GroupChatSummary\\\", \\\"plugin debug\\\", \\\"langbot test\\\".\",\n \"references\": [\n \"references/test-env-setup.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-skills-maintenance\",\n \"name\": \"langbot-skills-maintenance\",\n \"description\": \"Maintain the langbot-skills repository with low duplication. Use when adding, editing, or auditing LangBot skills, references, cases, troubleshooting entries, indexes, or periodic entropy-control checks for this skills repository.\",\n \"references\": [\n \"references/curation-workflow.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-space-ops\",\n \"name\": \"langbot-space-ops\",\n \"description\": \"Browse and search the LangBot Space marketplaces (plugins, MCP servers, skills) through the Space MCP server. Use when an AI agent needs to discover LangBot extensions on space.langbot.app over MCP. Covers the /mcp endpoint, Personal Access Token (PAT) auth, the tool surface, and client configuration. Triggers on \\\"langbot space mcp\\\", \\\"search langbot plugins\\\", \\\"langbot marketplace mcp\\\", \\\"space.langbot.app mcp\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-testing\",\n \"name\": \"langbot-testing\",\n \"description\": \"Test LangBot WebUI and core product flows with an automated browser and backend logs. Use when validating the configured LangBot frontend, pipeline Debug Chat, model provider setup and test buttons, bot and knowledge-base UI flows, or troubleshooting failed LangBot end-to-end tests.\",\n \"references\": [\n \"references/agent-runner-qa-workflow.md\",\n \"references/agent-runner-release-gate.md\",\n \"references/dify-agent-runner.md\",\n \"references/langrag-knowledge-base.md\",\n \"references/local-agent-runner-coverage.md\",\n \"references/local-agent-runner.md\",\n \"references/mcp-stdio-testing.md\",\n \"references/model-provider-testing.md\",\n \"references/performance-reliability-testing.md\",\n \"references/pipeline-debug-chat.md\",\n \"references/plugin-e2e-smoke.md\",\n \"references/sandbox-skill-authoring.md\",\n \"references/skill-all-tool-acceptance.md\",\n \"references/troubleshooting.md\",\n \"references/web-ui-testing.md\",\n \"references/workspace-release-testing.md\"\n ],\n \"cases\": [\n \"acp-agent-runner-debug-chat\",\n \"agent-run-ledger-audit\",\n \"agent-runner-async-db-readiness\",\n \"agent-runner-behavior-matrix\",\n \"agent-runner-fixture-contract\",\n \"agent-runner-health-visibility\",\n \"agent-runner-ledger-concurrency\",\n \"agent-runner-ledger-contention\",\n \"agent-runner-ledger-invariants\",\n \"agent-runner-ledger-stress\",\n \"agent-runner-live-install\",\n \"agent-runner-qa-debug-chat\",\n \"agent-runner-release-preflight\",\n \"agent-runner-runtime-chaos\",\n \"bot-event-routing-product-flow\",\n \"box-mcp-heartbeat-recovery\",\n \"claude-code-agent-debug-chat\",\n \"codex-agent-debug-chat\",\n \"dify-agent-debug-chat\",\n \"langbot-fake-provider-debug-chat-cross-pipeline-isolation\",\n \"langbot-fake-provider-debug-chat-fault-recovery\",\n \"langbot-fake-provider-debug-chat-load\",\n \"langbot-fake-provider-debug-chat-slow-load\",\n \"langbot-fault-taxonomy-contract\",\n \"langbot-live-backend-latency\",\n \"langbot-live-backend-log-health\",\n \"langbot-live-control-plane-api\",\n \"langbot-overhead-accounting-contract\",\n \"langbot-space-debug-chat-concurrency-smoke\",\n \"langrag-kb-retrieve\",\n \"langrag-parser-golden-e2e\",\n \"langrag-sentinel-kb-discover\",\n \"local-agent-basic-debug-chat\",\n \"local-agent-combo-rag-compaction-tool-debug-chat\",\n \"local-agent-complex-coding-task-debug-chat\",\n \"local-agent-context-compaction-debug-chat\",\n \"local-agent-effective-prompt-debug-chat\",\n \"local-agent-model-fallback-before-first-chunk-debug-chat\",\n \"local-agent-multimodal-debug-chat\",\n \"local-agent-multitool-rag-compaction-debug-chat\",\n \"local-agent-nonstreaming-debug-chat\",\n \"local-agent-parallel-tools-rag-compaction-debug-chat\",\n \"local-agent-plugin-tool-call-debug-chat\",\n \"local-agent-rag-debug-chat\",\n \"local-agent-rag-multimodal-debug-chat\",\n \"local-agent-steering-debug-chat\",\n \"local-agent-streaming-post-commit-failure-debug-chat\",\n \"local-agent-tool-error-recovery-debug-chat\",\n \"local-agent-tool-loop-limit-debug-chat\",\n \"mcp-stdio-register\",\n \"mcp-stdio-tool-call\",\n \"pipeline-debug-chat\",\n \"pipeline-debug-chat-performance\",\n \"plugin-e2e-smoke\",\n \"provider-deepseek\",\n \"qa-plugin-smoke-live-install\",\n \"sandbox-skill-authoring-e2e\",\n \"sandbox-skill-authoring-edit-existing-e2e\",\n \"skill-discovery-via-mcp-gateway\",\n \"webui-login-state\",\n \"wizard-onebot-agent-runtime\",\n \"wizard-runner-marketplace-catalog\",\n \"wizard-scenario-routing\",\n \"workspace-compatibility-preflight\",\n \"workspace-plugin-pages-smoke\",\n \"workspace-repository-contracts\"\n ],\n \"case_summaries\": [\n {\n \"id\": \"acp-agent-runner-debug-chat\",\n \"title\": \"ACP Runner can answer through Debug Chat using real remote Claude\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"acp\",\n \"claude\",\n \"external-runner\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_ACP_RUNNER_PIPELINE_URL\",\n \"LANGBOT_ACP_RUNNER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"agent-run-ledger-audit\",\n \"title\": \"Persisted Runner run ledger passes end-to-end invariants\",\n \"mode\": \"probe\",\n \"area\": \"agent\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"ledger\",\n \"audit\",\n \"tools\"\n ],\n \"automation\": \"scripts/e2e/agent-run-ledger-audit.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\",\n \"metrics\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-async-db-readiness\",\n \"title\": \"Runner async DB readiness probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"async-db\",\n \"aiosqlite\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-async-db-readiness.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-behavior-matrix\",\n \"title\": \"Runner deterministic behavior matrix probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"deterministic-runner\",\n \"protocol\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-behavior-matrix.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-fixture-contract\",\n \"title\": \"QA Runner fixture contract probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"fixture\",\n \"deterministic-runner\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-fixture-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-health-visibility\",\n \"title\": \"Agent configuration shows whether its selected runner is usable\",\n \"mode\": \"agent-browser\",\n \"area\": \"agent\",\n \"type\": \"feature\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"agent\",\n \"productization\",\n \"release-gate\"\n ],\n \"automation\": \"scripts/e2e/agent-runner-health-visibility.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-concurrency\",\n \"title\": \"Runner run ledger concurrency and auth pytest probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"pytest\",\n \"run-ledger\",\n \"concurrency\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-concurrency.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-contention\",\n \"title\": \"Runner ledger SQLite contention probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"stress\",\n \"ledger\",\n \"concurrency\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-contention.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-invariants\",\n \"title\": \"Runner ledger schema and status invariants probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"ledger\",\n \"invariant\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-invariants.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-stress\",\n \"title\": \"Runner ledger lightweight stress probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"stress\",\n \"ledger\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-stress.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-live-install\",\n \"title\": \"QA Runner package installs and registers in LangBot\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"plugin\",\n \"local-install\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/install-qa-plugin-smoke.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-qa-debug-chat\",\n \"title\": \"QA Runner returns deterministic output through Debug Chat\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"pipeline\",\n \"debug-chat\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"case:agent-runner-live-install\",\n \"node:scripts/e2e/ensure-qa-agent-runner-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_QA_RUNNER_PIPELINE_URL\",\n \"LANGBOT_QA_RUNNER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-release-preflight\",\n \"title\": \"Agent runner release gate preflight validates environment readiness\",\n \"mode\": \"agent-browser\",\n \"area\": \"release\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"release-gate\",\n \"preflight\",\n \"environment\"\n ],\n \"automation\": \"scripts/e2e/agent-runner-release-preflight.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-runtime-chaos\",\n \"title\": \"Runner SDK runtime chaos pytest probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"pytest\",\n \"runtime\",\n \"sdk\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-runtime-chaos.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"bot-event-routing-product-flow\",\n \"title\": \"Bot event routing and adapter input can be inspected from the WebUI\",\n \"mode\": \"agent-browser\",\n \"area\": \"bot\",\n \"type\": \"feature\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"bot\",\n \"eba\",\n \"event-routing\",\n \"productization\"\n ],\n \"automation\": \"scripts/e2e/bot-event-routing-product-flow.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"box-mcp-heartbeat-recovery\",\n \"title\": \"Box heartbeat restores an existing stdio MCP tool session\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"box\",\n \"mcp\",\n \"stdio\",\n \"reliability\",\n \"chaos\",\n \"fault-injection\"\n ],\n \"automation\": \"skills/langbot-testing/probes/box-mcp-heartbeat-recovery.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"metrics\",\n \"api_diagnostic\",\n \"resource_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"claude-code-agent-debug-chat\",\n \"title\": \"Claude Code Agent can retrieve LangBot history through MCP\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"claude-code\",\n \"external-runner\",\n \"mcp\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"codex-agent-debug-chat\",\n \"title\": \"Codex Agent can retrieve LangBot history through MCP\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"codex\",\n \"external-runner\",\n \"mcp\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"dify-agent-debug-chat\",\n \"title\": \"Dify Runner returns a response through Pipeline Debug Chat\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"provider\",\n \"priority\": \"p2\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"dify\",\n \"agent-runner\",\n \"pipeline\"\n ],\n \"automation\": \"\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-cross-pipeline-isolation\",\n \"title\": \"LangBot Debug Chat fake-provider cross-pipeline isolation probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"reliability\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"isolation\",\n \"concurrency\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-cross-pipeline-isolation.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-cross-pipelines.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_A_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_A_NAME\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_B_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_B_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-fault-recovery\",\n \"title\": \"LangBot Debug Chat fake-provider fault recovery probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"chaos\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"fault-injection\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-load\",\n \"title\": \"LangBot Debug Chat controlled fake-provider load probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"load\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-slow-load\",\n \"title\": \"LangBot Debug Chat slow fake-provider load probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"slow-provider\",\n \"load\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fault-taxonomy-contract\",\n \"title\": \"LangBot fault taxonomy and cleanup contract\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": true,\n \"tags\": [\n \"reliability\",\n \"chaos\",\n \"contract\",\n \"synthetic\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-fault-taxonomy-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-backend-latency\",\n \"title\": \"LangBot live backend basic latency probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"live-backend\",\n \"latency\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-backend-latency.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-backend-log-health\",\n \"title\": \"LangBot live backend log health probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"reliability\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"live-backend\",\n \"backend-log\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-backend-log-health.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"backend_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-control-plane-api\",\n \"title\": \"LangBot live control-plane API probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"reliability\",\n \"live-backend\",\n \"control-plane\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-control-plane-api.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-overhead-accounting-contract\",\n \"title\": \"LangBot overhead accounting metrics contract\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": true,\n \"tags\": [\n \"performance\",\n \"metrics\",\n \"contract\",\n \"synthetic\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-overhead-accounting-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"resource_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-space-debug-chat-concurrency-smoke\",\n \"title\": \"LangBot Debug Chat real Space-provider concurrency smoke\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"space\",\n \"live-provider\",\n \"smoke\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_PIPELINE_URL\",\n \"LANGBOT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_MODEL_UUID\",\n \"LANGBOT_E2E_MODEL_UUID\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langrag-kb-retrieve\",\n \"title\": \"LangRAG knowledge base ingests and retrieves a sentinel document\",\n \"mode\": \"agent-browser\",\n \"area\": \"knowledge\",\n \"type\": \"feature\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"langrag\",\n \"knowledge\",\n \"rag\"\n ],\n \"automation\": \"scripts/e2e/langrag-kb-retrieve.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"langrag-parser-golden-e2e\",\n \"title\": \"LangRAG and GeneralParsers retrieve a structured golden document\",\n \"mode\": \"agent-browser\",\n \"area\": \"knowledge\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"golden\",\n \"e2e\",\n \"langrag\",\n \"parser\",\n \"knowledge\"\n ],\n \"automation\": \"\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"langrag-sentinel-kb-discover\",\n \"title\": \"Existing LangRAG sentinel knowledge base is discoverable\",\n \"mode\": \"probe\",\n \"area\": \"knowledge\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"langrag\",\n \"knowledge\",\n \"rag\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/ensure-langrag-sentinel-kb.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-basic-debug-chat\",\n \"title\": \"Local Agent Debug Chat returns a deterministic streaming response\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"streaming\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-combo-rag-compaction-tool-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and plugin tool result in one Debug Chat run\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-complex-coding-task-debug-chat\",\n \"title\": \"Local Agent completes a multi-file coding task with iterative verification\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"agentic\",\n \"coding\",\n \"sandbox\",\n \"tools\",\n \"long-running\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/reset-complex-agent-task.mjs\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_COMPLEX_AGENT_WORKSPACE\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\",\n \"metrics\"\n ]\n },\n {\n \"id\": \"local-agent-context-compaction-debug-chat\",\n \"title\": \"Local Agent compacts long Debug Chat history and preserves older facts\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"context\",\n \"compaction\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-effective-prompt-debug-chat\",\n \"title\": \"Local Agent consumes host effective prompt after PromptPreProcessing\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"prompt\",\n \"plugin\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-model-fallback-before-first-chunk-debug-chat\",\n \"title\": \"Local Agent falls back when the primary model fails before streaming starts\",\n \"mode\": \"probe\",\n \"area\": \"pipeline\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"fallback\",\n \"streaming\",\n \"fake-provider\",\n \"fault-injection\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"local-agent-multimodal-debug-chat\",\n \"title\": \"Local Agent Debug Chat preserves uploaded image input\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"multimodal\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-multitool-rag-compaction-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and two-step plugin tool loop\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"tool-loop\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-nonstreaming-debug-chat\",\n \"title\": \"Local Agent Debug Chat returns a deterministic response with UI streaming disabled\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"non-streaming\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-parallel-tools-rag-compaction-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and parallel plugin tools\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"parallel-tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-plugin-tool-call-debug-chat\",\n \"title\": \"Local Agent can call a plugin-provided tool\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"plugin\",\n \"tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-rag-debug-chat\",\n \"title\": \"Local Agent Debug Chat answers from a LangRAG knowledge base\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-rag-multimodal-debug-chat\",\n \"title\": \"Local Agent preserves image input while using RAG context\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"rag\",\n \n\n","skills/skills/langbot-deploy/SKILL.md":"---\nname: langbot-deploy\ndescription: Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \"deploy langbot\", \"langbot docker\", \"langbot compose\", \"langbot kubernetes\", \"langbot config.yaml\", \"langbot box runtime\", \"langbot global api key\".\n---\n\n# LangBot Deployment & Configuration\n\nCovers running LangBot in production. For development see `langbot-dev`.\n\nRead https://langbot.app/llms.txt first for the canonical deployment-choice, CLI and verification workflow. The summary below applies when selecting hosting.\n\n## Deployment choice and optional managed hosting\n\nRespect the user's deployment preference and reuse an existing instance when available. Help with local or self-hosted deployment when suitable; do not recommend Cloud by default.\n\nOffer [LangBot Cloud](https://space.langbot.app/cloud) when the user wants managed hosting without maintaining a server, the local environment is unsuitable, or a deployment problem remains difficult after reasonable troubleshooting. Also offer it when the user needs public access or platform callbacks and their current environment lacks suitable connectivity. Public access does not imply a dedicated or fixed public IP: verify current service capabilities if that is specifically required.\n\nIf the user wants to support LangBot's continued development and open-source maintenance, mention the official hosted service as an optional way to do so, without pressure or repeated appeals. Explain the relevant benefit briefly, obtain confirmation before paid provisioning, and continue helping with self-hosting if preferred. Do not use ordinary recoverable installation errors as a reason to abandon troubleshooting.\n\n## Docker Compose (recommended)\n\n```bash\ngit clone https://github.com/langbot-app/LangBot\ncd LangBot/docker\n\n# Full stack (sandbox/Box + stdio MCP hosting + skill add/edit enabled)\ndocker compose --profile all up\n\n# Basic (no Box runtime)\ndocker compose up\n```\n\nThe `all` / `box` profile starts three services:\n\n- `langbot` — main app, serves API + UI on `:5300`.\n- `langbot_plugin_runtime` — plugin runtime (control `:5400`, debug `:5401`).\n- `langbot_box` — Box sandbox runtime (`:5410`). Uses the host Docker socket to\n spawn sandbox containers, so the **Box root host path and in-container path\n must be identical** (`BOX__LOCAL__HOST_ROOT=${LANGBOT_BOX_ROOT:-${PWD}/data/box}`).\n OSS allows its RPC and managed-process relay to run without a token when both\n sides leave `LANGBOT_BOX_CONTROL_TOKEN` unset. For an exposed endpoint, set\n the same value of at least 32 non-whitespace characters in both the LangBot\n and Box containers. Generate it once with `openssl rand -hex 32`; never put\n it in `box.runtime.endpoint` or commit it to config.\n\nA Compose deployment may optionally set\n`LANGBOT_PLUGIN_RUNTIME_CONTROL_TOKEN` on both `langbot` and\n`langbot_plugin_runtime` when port 5400 needs shared-secret protection. OSS\ndefaults to leaving it unset on both sides. If enabled, generate one value with\n`openssl rand -hex 32`; configuring only one side causes the control connection\nto fail. Kubernetes may use the `langbot-plugin-runtime-control` Secret shown in\n`docker/kubernetes.yaml`.\n\nWith Box off, the dashboard/skills list stays visible (read-only) but sandbox\ntools, skill add/edit, and stdio MCP are disabled. Set `box.enabled: false`\n(or `BOX__ENABLED=false`) to match.\n\n## Kubernetes\n\nSee `docker/kubernetes.yaml` and the deployment guide at\nhttps://langbot.app/docs. `docker/deploy-k8s-test.sh` is a test helper.\n\n## config.yaml (generated at `data/config.yaml` on first run)\n\nTop-level sections: `api`, `system`, `command`, `concurrency`, `proxy`,\n`database`, `vdb`, `storage`, `plugin`, `monitoring`, `box`, `space`.\n\nKey settings:\n\n| Key | Meaning |\n| --- | --- |\n| `api.port` | HTTP API + UI port (default 5300) |\n| `api.global_api_key` | **Global API key** for the HTTP API + MCP server. Non-empty = accepted with no login/DB record; no `lbk_` prefix required. Empty = disabled. Plaintext — trusted/internal only, serve over HTTPS. |\n| `plugin.runtime_ws_url` | Standalone plugin runtime WS URL (e.g. `ws://langbot_plugin_runtime:5400/control/ws`) |\n| `box.enabled` | Master switch for the Box sandbox runtime |\n| `box.backend` | `local` (Docker/nsjail autopick) / `docker` / `nsjail` / `e2b` / explicit unsafe `host`; env override `BOX__BACKEND` |\n| `box.runtime.endpoint` | External Box runtime URL (e.g. `ws://127.0.0.1:5410`); empty = local auto-managed |\n\nMany keys have `ENV__SUBKEY` overrides (e.g. `BOX__BACKEND`, `BOX__ENABLED`).\n\n## Runtimes & flags\n\n- LangBot started directly spawns the plugin runtime over **stdio**.\n- In containers it connects to a standalone runtime over **WebSocket**; start\n with `--standalone-runtime`.\n- Box has a parallel `--standalone-box` flag; the Docker box host is\n `langbot_box:5410`.\n- `box.backend: host` runs commands directly as the Box Runtime system user.\n It is never auto-selected, provides no sandbox isolation, and is only for\n trusted local development. A WebSocket-controlled host backend requires\n `LANGBOT_BOX_CONTROL_TOKEN`; local stdio control is allowed.\n\n## Global API key — enabling for agents/automation\n\n```yaml\n# data/config.yaml\napi:\n port: 5300\n global_api_key: 'a-strong-secret' # empty disables it\n```\n\nThis key authenticates both the HTTP API and the MCP server (`/mcp`) without a\nlogin session. See `langbot-mcp-ops` for using it, and `docs/API_KEY_AUTH.md`.\n\n## Pitfalls\n\n- \"No supported sandbox backend (Docker / nsjail / E2B)\" with Docker running\n usually means the user isn't in the `docker` group →\n `sudo usermod -aG docker ` and restart in a new shell.\n- Do not use `box.backend: host` as a production fallback. It cannot enforce\n image, filesystem, network, PID, CPU, memory, or storage isolation.\n- Box root host/container path mismatch breaks sandbox container creation.\n- Don't commit a non-empty `api.global_api_key` to version control.\n","skills/skills/langbot-dev/SKILL.md":"---\nname: langbot-dev\ndescription: Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \"langbot backend\", \"langbot dev\", \"langbot api\", \"add langbot endpoint\", \"langbot migration\".\n---\n\n# LangBot Core Development\n\nThis skill covers developing the LangBot core (the main repo), distinct from\nplugin development (see `langbot-plugin-dev`) and deployment (`langbot-deploy`).\n\n## Stack\n\n- **Backend**: Python `>=3.11,<4.0`, deps via `uv`. Framework: **Quart** (async\n Flask). Serves the HTTP API + pre-built web UI on `http://127.0.0.1:5300`.\n- **Frontend** (`web/`): **Vite + React Router 7 + shadcn/ui + Tailwind**,\n managed by `pnpm`. Dev server on `:3000`. (NOT Next.js — `dev` script is `vite`.)\n\n## Dev environment\n\n```bash\n# Backend\npip install uv\nuv sync --dev\nuv run main.py # API + UI on http://127.0.0.1:5300\n\n# Frontend (separate terminal)\ncd web\ncp .env.example .env\npnpm install\npnpm dev # http://127.0.0.1:3000 (reads VITE_API_BASE_URL)\n\n# Lint/format hooks (CI runs the same checks)\nuv run pre-commit install\n```\n\nFirst run generates `data/config.yaml`; DB defaults to SQLite (PostgreSQL\nsupported). Migrations run automatically on startup.\n\n## Repo layout (key paths)\n\n```\nsrc/langbot/\n├── __main__.py # entrypoint, CLI flags (--standalone-runtime/-box/--debug)\n├── pkg/\n│ ├── api/\n│ │ ├── http/ # Quart controllers + services\n│ │ │ ├── controller/groups/ # route groups (@group.group_class)\n│ │ │ └── service/ # business logic (called by controllers AND MCP)\n│ │ └── mcp/ # MCP server (server.py = tools, mount.py = ASGI dispatch)\n│ ├── core/ # app bootstrap, stages, task manager\n│ ├── platform/ provider/ pipeline/ plugin/ box/ skill/ rag/ vector/\n│ ├── command/ persistence/ storage/ config/ entity/ telemetry/\n│ └── templates/config.yaml # config template (top-level: api, system, plugin, box, space...)\n├── web/ # Vite SPA\n└── docker/ # compose deployment\n```\n\n## HTTP API auth model\n\nRoute auth is declared per-route via `AuthType` in\n`pkg/api/http/controller/group.py`:\n\n- `NONE` — public.\n- `USER_TOKEN` — web UI JWT (`Authorization: Bearer `).\n- `API_KEY` — `X-API-Key` or `Authorization: Bearer `.\n- `USER_TOKEN_OR_API_KEY` — either.\n\nAuthenticated routes receive an immutable `RequestContext` containing the\nprincipal, authorized Workspace membership, fixed-role permissions, instance,\nrequest id, and placement generation. A browser's `X-Workspace-Id` is only a\nselector and is always checked against the Account membership. Tenant services\nmust accept this context (or an explicit trusted execution context) and fail\nclosed when it is absent.\n\nAPI-key authentication accepts:\n\n1. the **global key** from `config.yaml` `api.global_api_key` only for a\n community instance with exactly one local Workspace, then\n2. **web-UI keys** whose one-time `lbk_` secret is stored only as a hash and is\n bound to one Workspace, explicit scopes, status, and optional expiry.\n\nAn API key derives its Workspace from the key record and ignores a caller's\nWorkspace selector. Public Bot/Webhook routes similarly derive Workspace from\nthe opaque owning resource rather than a header.\n\nRoute groups self-register via `@group.group_class(name, path)` and are\ndiscovered by `importutil.import_modules_in_pkg`.\n\n## Adding an API endpoint\n\n1. Add/extend a controller in `pkg/api/http/controller/groups/` and the matching\n service method in `pkg/api/http/service/`.\n2. Pick the right `AuthType`.\n3. **If the endpoint should be agent-accessible, add/adjust the matching MCP tool\n in `pkg/api/mcp/server.py` and update the `langbot-mcp-ops` skill.** API and\n MCP surface must stay aligned (see `AGENTS.md`).\n4. If `lbctl` uses the route or capability, check the separate\n [`langbot-cli`](https://github.com/langbot-app/langbot-cli) client for compatibility.\n5. Update `docs/service-api-openapi.json` if you maintain the OpenAPI overview.\n\n## Database migrations (Alembic)\n\nSingle migration set supports SQLite + PostgreSQL. Files in\n`src/langbot/pkg/persistence/alembic/versions/`.\n\n```bash\n# From project root (needs data/config.yaml)\nuv run python -m langbot.pkg.persistence.alembic_runner autogenerate \"description\"\n```\n\n## Standards\n\n- All code comments/docstrings in **English**; user-facing strings need **i18n**\n (`en_US` + `zh_Hans` minimum, `ja_JP` where present).\n- Consider toC and toB compatibility + security.\n- Commit format: `(): ` (feat/fix/docs/refactor/...).\n\n## Tests\n\n```bash\nuv run pytest tests/unit_tests -q # unit tests\nuv run pytest tests/unit_tests/api -q # API service tests\nuv run python tests/manual/mcp_smoke.py # MCP server e2e smoke\n```\n\n## See also\n\n- `langbot-plugin-dev` — plugin SDK / runtime development.\n- `langbot-testing` — WebUI/e2e QA harness (`bin/lbs`).\n- `langbot-deploy` — Docker/compose deployment + config.\n- `langbot-mcp-ops` — operating the LangBot MCP server.\n- [`langbot-cli`](https://github.com/langbot-app/langbot-cli) — `lbctl`, a standalone Service API client for managing running Workspaces.\n","skills/skills/langbot-eba-adapter-dev/SKILL.md":"---\nname: langbot-eba-adapter-dev\ndescription: Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing.\n---\n\n# LangBot EBA Adapter Development\n\nUse this skill when implementing or reviewing a LangBot platform adapter under the Event-Based Agents architecture.\n\n## Controlling a running instance via MCP\n\nBeyond writing code, you can **drive a live LangBot instance over MCP** — no raw\nHTTP needed. Two MCP servers exist (both reuse existing API keys; see `AGENTS.md`):\n\n- **LangBot instance** — `http://:5300/mcp` (auth: web-UI `lbk_` key or the\n `api.global_api_key` from `config.yaml`). Manage bots, pipelines, models,\n knowledge bases, and skills. See the **`langbot-mcp-ops`** skill.\n- **LangBot Space marketplace** — `https://space.langbot.app/mcp` (auth: Personal\n Access Token). Search plugins / MCP servers / skills. See the\n **`langbot-space-ops`** skill.\n\n> Any change to an agent-accessible HTTP API endpoint must keep the matching MCP\n> tool and these skills in sync.\n\n## Core Rule\n\nDo not let platform-native event or message shapes leak into LangBot's common path. Each adapter must convert incoming SDK objects into unified EBA entities before dispatch:\n\n- Events: `langbot_plugin.api.entities.builtin.platform.events`\n- Message chains: `langbot_plugin.api.entities.builtin.platform.message.MessageChain`\n- Users/groups/members: `langbot_plugin.api.entities.builtin.platform.entities`\n- Raw platform objects may remain only in `source_platform_object` for debugging or platform-specific escape hatches.\n\n## Start Here\n\n1. Read the EBA design docs in `LangBot/docs/event-based-agents/`.\n2. Read the architecture-level acceptance checklist before writing or validating code:\n - `LangBot/docs/event-based-agents/adapters/acceptance-checklist.md`\n3. Read the current reference adapter before writing code. Prefer Telegram first:\n - `LangBot/src/langbot/pkg/platform/adapters/telegram/`\n - `LangBot/docs/event-based-agents/adapters/telegram.md`\n4. Read the legacy source adapter for the target platform:\n - `LangBot/src/langbot/pkg/platform/sources/.py`\n - `LangBot/src/langbot/pkg/platform/sources/.yaml`\n5. Inspect SDK entity definitions in `langbot-plugin-sdk/src/langbot_plugin/api/entities/builtin/platform/`.\n6. Search before assuming APIs. Platform SDKs change often.\n\n## Adapter Layout\n\nCreate one directory per adapter:\n\n```text\nLangBot/src/langbot/pkg/platform/adapters//\n├── __init__.py\n├── adapter.py\n├── api_impl.py\n├── event_converter.py\n├── manifest.yaml\n├── message_converter.py\n├── platform_api.py\n├── types.py\n└── .svg\n```\n\nAdd optional helpers such as `voice.py` only when the platform has a real domain-specific surface.\n\nEnsure `pyproject.toml` package data includes adapter assets:\n\n```toml\npackage-data = { \"langbot\" = [\"templates/**\", \"pkg/platform/sources/*\", \"pkg/platform/adapters/**\", ...] }\n```\n\n## Implementation Checklist\n\n- `manifest.yaml` declares `metadata.name`, config schema, supported events, common APIs, and platform-specific APIs.\n- `adapter.py` creates the platform client, subscribes to native events, filters self/bot loops where appropriate, calls `event_converter.target2yiri(...)`, then dispatches the EBA event.\n- `event_converter.py` maps native events to EBA event classes such as `MessageReceivedEvent`, `MessageEditedEvent`, `MessageDeletedEvent`, `MessageReactionEvent`, `MemberJoinedEvent`, `BotInvitedToGroupEvent`, and `PlatformSpecificEvent`.\n- `message_converter.py` maps native messages to `MessageChain`, and maps `MessageChain` back to the platform send format.\n- `api_impl.py` implements common EBA APIs: send, reply, edit, delete, forward, user/group/member lookup, moderation, upload/file URL, leave group.\n- `platform_api.py` keeps platform-specific calls behind `call_platform_api(action, params)`.\n- Unsupported common APIs must raise explicit SDK platform errors such as `NotSupportedError`; do not silently no-op.\n- Destructive APIs such as kick, ban, leave, delete, or moderation must be gated in live tests and documented.\n\n## Conversion Contract\n\nFor message events, the common shape should look like this regardless of platform:\n\n```python\nplatform_events.MessageReceivedEvent(\n type=\"message.received\",\n adapter_name=\"\",\n message_id=,\n message_chain=platform_message.MessageChain([...]),\n sender=platform_entities.User(...),\n chat_type=platform_entities.ChatType.PRIVATE or ChatType.GROUP,\n chat_id=,\n group=platform_entities.UserGroup(...) or None,\n source_platform_object=,\n)\n```\n\nMessage content should use common components:\n\n- `Source` for original message id/time when available.\n- `Plain` for text.\n- `At` / `AtAll` for mentions.\n- `Image`, `Voice`, `File` for media.\n- `Forward` only when the platform can represent or emulate it safely.\n\nIf a platform event cannot cleanly map to a common event, emit `PlatformSpecificEvent` with a compact `action` and structured `data`.\n\n## Unit Tests\n\nAdd focused tests under `LangBot/tests/unit_tests/platform/test__eba_adapter.py`.\n\nCover at least:\n\n- Manifest supported events match adapter `supported_events()`.\n- Manifest supported APIs match adapter `supported_apis()`.\n- Platform API map matches manifest actions.\n- Dispatcher chooses the most specific EBA listener.\n- Message converter maps every supported common component both directions where possible:\n - `Source`\n - `Plain`\n - `At`\n - `AtAll`\n - `Image`\n - `Voice`\n - `File`\n - `Quote`\n - `Face`\n - `Forward`\n - `Unknown`\n - mixed chains preserving order\n- Event converter maps message received/edited/deleted/reaction, raw uncached gateway events, member events, and bot join/leave events.\n- Send/reply methods pass correct platform kwargs and return `MessageResult`.\n\nRun the existing reference adapter tests too:\n\n```bash\ncd LangBot\nuv run pytest tests/unit_tests/platform/test__eba_adapter.py tests/unit_tests/platform/test_telegram_eba_adapter.py\nuv run python -m py_compile tests/e2e/live__eba_probe.py\ngit diff --check\n```\n\n## Live Test Workflow\n\nDirect adapter live probes are useful diagnostics, but they are not sufficient acceptance evidence for EBA. Treat `tests/e2e/live__eba_probe.py` as an auxiliary tool only. The final adapter record must distinguish:\n\n- `plugin-e2e-ui`: real SDK plugin through standalone runtime, LangBot core, adapter, and a real/simulator UI action. This can mark an inbound UI item complete.\n- `plugin-e2e-protocol`: real SDK plugin through standalone runtime, LangBot core, adapter, and a protocol-boundary injected event. This is useful evidence but must not be claimed as UI coverage.\n- `plugin-e2e-outbound`: real SDK plugin calls an API and the bot output is visible in the real/simulator UI. This can mark send/API coverage complete.\n- `adapter-live`: direct adapter probe connected to a real/simulator endpoint. This is auxiliary only.\n- `unit`: mocked conversion/API-shape coverage. This is auxiliary only.\n- `not-supported`: platform protocol or SDK has no equivalent. Must include the reason.\n- `blocked`: intended capability could not be verified. This is not complete.\n\nWrite a live probe in `LangBot/tests/e2e/live__eba_probe.py`. It should:\n\n1. Read token/client ids from environment variables or CLI args.\n2. Start the adapter directly.\n3. Register an EBA listener and write JSONL evidence to `LangBot/data/temp/`.\n4. Wait for a real user/platform event instead of fabricating the entrypoint.\n5. Exercise common APIs and `call_platform_api` actions.\n6. Observe returned gateway events for edit/delete/reaction/member/bot lifecycle where available.\n7. Print a summary containing passed, failed, skipped, and observed event types.\n8. Redact or avoid printing secrets.\n9. Keep destructive operations behind flags and run them last.\n\nUse Computer Use when the user asks for real platform end-to-end coverage. Actually send messages/click reactions in the platform UI or otherwise trigger real user-side events; do not replace that with unit tests.\n\nFor media/component acceptance, keep the direction and trigger source explicit:\n\n- Real inbound media only counts when a human-side platform UI or simulator UI sends the image/file/voice to the bot and the plugin JSONL records the corresponding common component.\n- Bot outbound media only proves `send_message`/adapter send conversion. It does not prove inbound conversion.\n- Protocol-boundary injection, such as sending a OneBot event directly into a reverse WebSocket adapter, is useful and should be labelled `plugin-e2e-protocol`, but it must not be reported as UI-level end-to-end media upload.\n- If the UI cannot send or upload the media, record the item as `blocked` with the exact client/simulator limitation.\n\n## Standalone Runtime + Plugin Test\n\nWhen validating the whole LangBot EBA path, test with the SDK standalone runtime and a real test plugin. This is the required acceptance path; direct adapter calls do not prove the EBA architecture path.\n\nThe required path is:\n\n```text\nReal platform / simulator UI\n -> platform SDK native event\n -> adapter event converter\n -> unified EBA event/entity/message types\n -> LangBot core event dispatch\n -> standalone SDK runtime\n -> real test plugin listener\n -> plugin calls platform APIs through SDK\n -> LangBot core API dispatch\n -> adapter API implementation\n -> real platform / simulator UI\n```\n\nTypical shape:\n\n```bash\n# Terminal 1, SDK repo\ncd langbot-plugin-sdk\nuv run python -m langbot_plugin.cli.__init__ rt \\\n --debug-only \\\n --ws-control-port 5400 \\\n --ws-debug-port 5401 \\\n --skip-deps-check\n\n# Terminal 2, LangBot repo\ncd LangBot\nexport PYTHONPATH=/absolute/path/to/langbot-plugin-sdk/src:${PYTHONPATH:-}\nuv run main.py --standalone-runtime\n\n# Terminal 3, plugin directory\nexport DEBUG_RUNTIME_WS_URL=ws://127.0.0.1:5401/plugin/ws\nexport EBA_PROBE_LOG=/absolute/path/to/LangBot/data/temp/_eba_plugin_probe.jsonl\nexport EBA_PROBE_API=1\nexport EBA_PROBE_COMPONENT_SWEEP=1\nexport EBA_PROBE_PLATFORM_API=1\nuv --project /absolute/path/to/langbot-plugin-sdk run python -m langbot_plugin.cli.__init__ run\n```\n\nUse an EBA probe plugin that subscribes to all relevant EBA event classes and runs SDK API calls after the first `MessageReceived`.\n\nThe plugin evidence should be JSONL and include:\n\n- event class and `event.type`\n- adapter name\n- chat type and chat ID\n- sender/user/group IDs with secrets redacted\n- `bot_uuid` and `adapter_name`, proving LangBot filled common routing fields before plugin dispatch\n- received `message_chain` component list\n- API action name, input summary, result or error\n- unsupported or blocked reason when an item is skipped\n\nFor full adapter acceptance, enable both probe sweeps:\n\n- `EBA_PROBE_COMPONENT_SWEEP=1` sends the required outbound message components through `send_message`.\n- `EBA_PROBE_PLATFORM_API=1` calls common safe APIs plus selected `call_platform_api` actions for the adapter.\n\nThe SDK must support `plugin.call_platform_api(bot_uuid, action, params)` for platform-specific acceptance. If the SDK cannot call a platform-specific action from the plugin, the adapter cannot be fully accepted even if direct adapter probes pass.\n\n## Required EBA Acceptance Coverage\n\nBefore marking an adapter migrated, fill out an adapter record against `LangBot/docs/event-based-agents/adapters/acceptance-checklist.md`.\n\nAt minimum, the record must cover these categories:\n\n- Message receive component tests through `plugin-e2e-ui`: `Source`, `Plain`, `At`, `AtAll`, `Image`, `Voice`, `File`, `Quote`, `Face`, `Forward`, `Unknown`, and mixed chains where the platform supports them. Protocol-only receive evidence must be labelled `plugin-e2e-protocol`.\n- Message send component tests through `plugin-e2e-outbound`: `Plain`, `At`, `AtAll`, `Image`, `Voice`, `File`, `Quote`, `Face`, `Forward`, and mixed chains where the platform supports them.\n- Every event declared in `manifest.yaml -> spec.supported_events`.\n- Every common API declared in `manifest.yaml -> spec.supported_apis.required` and `optional`.\n- Every action declared in `manifest.yaml -> spec.platform_specific_apis`.\n- Compatibility tests for manifest declarations, legacy message listener fallback, EBA listener specificity, bot self-message filtering, and `source_platform_object` reply/debug behavior.\n\nDo not declare an event or API in the manifest unless it has an implementation path and an acceptance entry. If a platform or simulator lacks a capability, document it as `not-supported` or `blocked` rather than silently omitting the test.\n\n## Common Pitfalls\n\n- `get_bots()` may return bot dictionaries, not UUID strings. Probe plugins should select an enabled dict and pass `bot[\"uuid\"]` to `get_bot_info()` and `send_message()`.\n- Make sure the probe subscribes to every event you claim to verify. Missing `MessageDeleted` subscription can make a working adapter look untested.\n- Some platforms emit both cached and raw gateway events, producing duplicate evidence for delete/reaction. Count this explicitly; do not treat duplicates as failure unless semantics differ.\n- Self-message filtering is platform-specific. Filter bot-originated `message.received` loops, but do not accidentally filter edit/delete events needed for bot-owned API probes.\n- Reaction events may be filtered for bot self reactions. To test user reaction add/remove, use real UI interaction or a real user token path if permitted.\n- File uploads usually happen as message attachments. A standalone `upload_file` API may need to be `NotSupportedError`.\n- Live probes should not leak bot tokens through command output, logs, docs, or final answers.\n- Discord requires privileged intents for message content and members. Missing intents can look like converter bugs.\n- Telegram Bot API exposes only limited member lists; document capability gaps.\n- Do not mark moderation APIs verified unless they ran against a disposable target member/bot.\n- If `leave_group` is tested, run it last because the test bot will be removed from the server/group.\n- Restore local LangBot DB/test state after live runs if you enabled temporary bots or changed plugin settings.\n\n## Documentation Record\n\nAdd or update `LangBot/docs/event-based-agents/adapters/.md` in the same style as Telegram:\n\n- Status and adapter directory.\n- Configuration table matching manifest fields.\n- Supported EBA event list.\n- Common API table with support and limitations.\n- `call_platform_api` action list.\n- Receive component table with evidence level per component.\n- Send component table with evidence level per component.\n- Event table with evidence level per event.\n- Common API table with evidence level per API.\n- Platform-specific API table with evidence level per action.\n- Live test record with exact date, endpoint/simulator, standalone runtime command, test plugin path/name, JSONL evidence path, channel/group type, observed events, APIs exercised, destructive operations, and skipped items.\n\nBe honest. Put untested or skipped APIs in the document with the reason. Do not imply full parity when a platform cannot provide the same information density.\n\n## Before Finishing\n\n- Run unit tests and compile the live probe.\n- Run the standalone runtime plugin E2E path for every required acceptance item that the platform supports.\n- Run `git diff --check`.\n- Summarize live JSONL evidence by event type.\n- Stop all long-running runtimes and probes.\n- Confirm no secrets are staged.\n- Leave unrelated untracked files alone.\n","skills/skills/langbot-env-setup/SKILL.md":"---\nname: langbot-env-setup\ndescription: Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths.\n---\n\n# LangBot Environment Setup\n\nUse this skill when a task needs LangBot to be in a testable state before product testing or development verification.\n\n## Routing\n\n- **Shared local variables**: read `../.env` before using URL, path, browser profile, or proxy defaults.\n- **Always start here**: read `references/browser-access-selection.md` to choose the browser-control path.\n- **LangBot service checks and startup**: read `references/service-startup.md`.\n- **Computer Use available**: read `references/computer-use.md`. This path usually needs less browser/MCP setup.\n- **No Computer Use, browser automation required**: read `references/playwright-mcp.md`.\n- **GitHub OAuth or persisted login profile**: read `references/oauth-browser-profile.md`.\n- **WSL-specific notes**: read `references/wsl-notes.md` only when running under WSL.\n- **Proxy setup**: read `references/proxy.md` when external login, model provider tests, or package downloads time out.\n- **Headless-only automation**: use only after a profile already contains a valid LangBot login. Do not ask the agent to enter GitHub credentials or 2FA.\n\n## Rules\n\n- Never handle the user's GitHub password, passkey, recovery code, or 2FA secret.\n- For OAuth login, open a visible browser and let the user complete the credential steps.\n- Reuse a fixed browser profile path so the agent can later access the logged-in LangBot session.\n- Keep environment-specific paths and commands in `references/`, not in this file.\n- Treat environment setup as complete only after the target LangBot services are reachable and the browser profile can access the WebUI.\n","skills/skills/langbot-env-setup/references/browser-access-selection.md":"# Browser Access Selection\n\nChoose the lightest browser-control path that can complete the task.\n\n## Decision Order\n\n1. If Codex Computer Use, Claude Computer Use, or another visible browser-control tool is available, use `computer-use.md`.\n2. If no computer-control tool is available but Playwright MCP is available, use `playwright-mcp.md`.\n3. If the browser session must survive restarts or OAuth login is required, also use `oauth-browser-profile.md`.\n4. If running under WSL, add `wsl-notes.md`.\n5. If external sites or model providers time out, add `proxy.md`.\n\n## Principle\n\nComputer Use and Playwright MCP are alternative browser-control paths. Both still need LangBot services to be reachable, so service checks stay in `service-startup.md`.\n","skills/skills/langbot-env-setup/references/computer-use.md":"# Computer Use Browser Path\n\nUse this path when Codex Computer Use, Claude Computer Use, or another agent-visible browser-control capability is available.\n\n## Why This Path Is Simpler\n\nComputer Use can interact with a visible browser directly, so it usually does not need Playwright MCP configuration or a separate MCP browser bridge.\n\n## Workflow\n\n1. Verify LangBot backend/frontend with `service-startup.md`.\n2. Open the WebUI in the controlled browser.\n3. If login is needed, let the user complete GitHub OAuth. Never handle credentials or 2FA.\n4. Keep the browser/profile available for later testing.\n5. Hand off to `langbot-testing` after the page shows the logged-in WebUI.\n\n## Still Required\n\n- Proxy may still be needed for GitHub OAuth or model provider tests. Use `proxy.md`.\n- Persisted profile details may still matter if the computer-control browser is restarted. Use `oauth-browser-profile.md` if login state must survive.\n","skills/skills/langbot-env-setup/references/oauth-browser-profile.md":"# OAuth Browser Profile\n\nUse this reference when LangBot or LangBot Space needs GitHub OAuth login and the agent must reuse the authenticated browser state later.\n\nRead `skills/.env` first for `LANGBOT_BACKEND_URL`, `LANGBOT_FRONTEND_URL`, `LANGBOT_BROWSER_PROFILE`, `LANGBOT_CHROMIUM_EXECUTABLE`, and proxy defaults.\n\n## Rules\n\n- Never handle the user's GitHub password, passkey, recovery code, or 2FA secret.\n- Open a visible browser and let the user complete credential steps.\n- Reuse a fixed browser profile path.\n- Do not print token values. It is acceptable to report localStorage key names.\n\n## Manual Visible Login Flow\n\n1. Verify LangBot backend is reachable with `service-startup.md`.\n2. Launch a visible Chromium window with the persistent profile:\n\n```bash\nsetsid \"$LANGBOT_CHROMIUM_EXECUTABLE\" \\\n --no-sandbox \\\n --ozone-platform=x11 \\\n --user-data-dir=\"$LANGBOT_BROWSER_PROFILE\" \\\n --proxy-server=\"$LANGBOT_PROXY_SOCKS\" \\\n --proxy-bypass-list=\"$LANGBOT_NO_PROXY\" \\\n \"$LANGBOT_BACKEND_URL/login\" \\\n >/tmp/langbot-visible-chrome.log 2>&1 < /dev/null &\n```\n\n3. The user completes:\n\n```text\nLogin with Space -> Login with GitHub -> GitHub credentials / 2FA -> authorize\n```\n\n4. The agent can then reuse the same profile for automated checks.\n\n## Expected Successful State\n\nAfter login, LangBot should redirect away from `/login`, for example to a `/home/...` URL on the selected origin.\n\nExpected visible signals:\n\n```text\nLangBot\nDashboard\nHome\nBots\nPipelines\nKnowledge\nExtensions\n```\n\nExpected localStorage key names:\n\n```text\ntoken\nuserEmail\nlangbot_language\n```\n\nIf the user logged in on one origin but `LANGBOT_FRONTEND_URL` still shows `/login`, copy only the auth state needed between origins. Do not print token values.\n","skills/skills/langbot-env-setup/references/playwright-mcp.md":"# Playwright MCP Browser Path\n\nUse this path when the agent needs browser automation but no Computer Use browser-control path is available.\n\n## Known Paths\n\n- Persistent browser profile: `LANGBOT_BROWSER_PROFILE` from `skills/.env.local`\n- Chromium executable: `LANGBOT_CHROMIUM_EXECUTABLE` from `skills/.env.local`\n- Codex MCP config: `$CODEX_HOME/config.toml` or the config path used by the active agent.\n\n## MCP Config\n\nKeep the profile path fixed so the agent can reuse authenticated state.\n\n```toml\n[mcp_servers.playwright]\ncommand = \"npx\"\nargs = [\"-y\", \"@playwright/mcp@latest\", \"--no-sandbox\", \"--executable-path\", \"\", \"--proxy-server\", \"\", \"--proxy-bypass\", \"localhost,127.0.0.1\", \"--user-data-dir\", \"\"]\n```\n\nAfter changing MCP config, restart Codex so the MCP server is relaunched with the new args.\n\n## Visible Login\n\nFor OAuth login, Playwright MCP's headless browser is not enough. Launch a visible browser with the same profile and let the user complete login. Use `oauth-browser-profile.md`.\n\n## Common Failures\n\n- MCP still uses old args after editing config: restart Codex or kill old `playwright-mcp` processes and restart the session.\n- Browser is headless during OAuth: use the visible login command from `oauth-browser-profile.md`.\n","skills/skills/langbot-env-setup/references/proxy.md":"# Proxy Setup\n\nUse this reference when GitHub OAuth, package installation, model provider tests, or external API calls time out.\n\nRead defaults from `skills/.env` first.\n\n## Standard Local Proxy\n\n```bash\nexport HTTP_PROXY=\"$LANGBOT_PROXY_HTTP\"\nexport HTTPS_PROXY=\"$LANGBOT_PROXY_HTTP\"\nexport ALL_PROXY=\"$LANGBOT_PROXY_SOCKS\"\nexport http_proxy=\"$LANGBOT_PROXY_HTTP\"\nexport https_proxy=\"$LANGBOT_PROXY_HTTP\"\nexport all_proxy=\"$LANGBOT_PROXY_SOCKS\"\nexport NO_PROXY=\"$LANGBOT_NO_PROXY\"\nexport no_proxy=\"$LANGBOT_NO_PROXY\"\n```\n\n## Rule\n\nKeep uppercase and lowercase proxy variables consistent. Different libraries read different names.\n\n## Checks\n\n```bash\nenv | rg -i '^(http|https|all|no)_?proxy='\ncurl -I --max-time 8 --proxy \"$LANGBOT_PROXY_SOCKS\" https://github.com\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL\"\n```\n","skills/skills/langbot-env-setup/references/service-startup.md":"# Service Startup\n\nUse this reference for LangBot backend/frontend readiness checks regardless of OS or browser-control method. Read `skills/.env` first and override those defaults with user-provided values or detected running services.\n\n## Variables\n\n- `LANGBOT_REPO`\n- `LANGBOT_WEB_REPO`\n- `LANGBOT_BACKEND_URL`\n- `LANGBOT_FRONTEND_URL`\n- `LANGBOT_DEV_FRONTEND_URL`\n\n## Backend\n\nStart LangBot from the backend repo:\n\n```bash\ncd \"$LANGBOT_REPO\"\nuv run main.py\n```\n\nHealthy startup includes:\n\n```text\nRunning on http://0.0.0.0:\nConnected to plugin runtime.\nPlugin langbot-team/LocalAgent initialized\n```\n\nQuick check:\n\n```bash\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL/login\"\n```\n\nIf `bin/lbs env doctor` reports that `LANGBOT_BACKEND_URL` has no TCP listener,\nthe backend is not running at the configured host and port. A reachable\nstandalone frontend on `LANGBOT_FRONTEND_URL` does not prove backend readiness.\n\nPrefer a visible terminal session while debugging backend startup. Detached\nbackground startup methods can hide early process exits in local agent runs; if\nyou use one, immediately verify both the process and the listener:\n\n```bash\nps -eo pid,cmd | rg 'main.py|uv run main|langbot'\nss -ltnp | rg ':5300'\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL/login\"\n```\n\n## Frontend\n\nStart the new frontend from the web repo:\n\n```bash\ncd \"$LANGBOT_WEB_REPO\"\nVITE_API_BASE_URL=\"$LANGBOT_BACKEND_URL\" pnpm dev --host 0.0.0.0\n```\n\nHealthy startup includes:\n\n```text\nLocal: \n```\n\nQuick check:\n\n```bash\ncurl -I --max-time 3 \"$LANGBOT_FRONTEND_URL\"\n```\n\nIf `VITE_API_BASE_URL` is missing, Vite still serves the page but frontend API\ncalls may go to the frontend port instead of the backend port. That produces\nfalse browser failures in login, wizard, pipeline, and Debug Chat cases.\n\n## Completion Signal\n\nEnvironment setup is not complete until the required frontend/backend URLs are reachable and the chosen browser-control path can open the WebUI.\n","skills/skills/langbot-env-setup/references/wsl-notes.md":"# WSL Notes\n\nUse this reference only for WSL-specific details. Do not put generic LangBot startup or browser-login steps here.\n\n## Network\n\nGitHub login and model provider calls may require proxy access from WSL.\n\nWorking proxy form:\n\n```bash\nsocks5://127.0.0.1:7890\n```\n\nBypass local LangBot:\n\n```bash\nlocalhost,127.0.0.1\n```\n\nQuick checks:\n\n```bash\ncurl -I --max-time 8 --proxy socks5h://127.0.0.1:7890 https://github.com\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL\"\n```\n\n## Visible Browser\n\nIf OAuth requires a visible browser, WSL must have a usable display path. If a visible Chromium launch fails, check the local WSL GUI/X11 setup before changing LangBot config.\n\n## Common Failures\n\n- `ERR_NETWORK_CHANGED` or GitHub timeout: browser is not using the SOCKS proxy.\n- LangBot connection refused: backend is not running or not reachable from WSL.\n- User cannot type credentials: browser is headless or not visible to the user.\n","skills/skills/langbot-mcp-ops/SKILL.md":"---\nname: langbot-mcp-ops\ndescription: Operate a LangBot instance through its built-in MCP (Model Context Protocol) server. Use when an AI agent needs to manage LangBot — list/create/update/delete bots, agents, pipelines, models, knowledge bases, MCP servers, and skills — over MCP instead of raw HTTP. Covers the /mcp endpoint, API-key auth (web-UI lbk_ keys and the config.yaml global key), the tool surface, and client configuration. Triggers on \"langbot mcp\", \"manage langbot via mcp\", \"langbot /mcp\", \"langbot mcp server\".\n---\n\n# LangBot MCP Operations\n\nLangBot exposes an **MCP server** so AI agents can manage an instance\nprogrammatically. It mirrors a curated subset of the HTTP service API.\n\n## Endpoint\n\n```\nhttp://:5300/mcp\n```\n\nTransport: **streamable HTTP** (stateless, JSON responses). Same host/port as\nthe web UI and HTTP API.\n\n## Authentication\n\nReuses the same API keys as the HTTP API. Send either header:\n\n```\nX-API-Key: \n# or\nAuthorization: Bearer \n```\n\nTwo kinds of key are accepted:\n\n1. **Web-UI key** — created in the web UI (sidebar → API Keys), prefixed `lbk_`.\n The secret is shown once; only its SHA-256 hash is stored. Each key is bound\n to one Workspace and has explicit scopes, status, optional expiry, and\n last-used metadata. The key determines the Workspace; callers cannot switch\n it with `X-Workspace-Id`.\n2. **Global API key** — set in `data/config.yaml` under `api.global_api_key`.\n Requires no login session and no DB record; does not need the `lbk_` prefix.\n It is accepted only by a community instance with exactly one local\n Workspace and is disabled for SaaS multi-Workspace operation. Leave empty to\n disable. See the `langbot-deploy` skill for config details.\n\nInvalid, revoked, or expired keys get `401 Unauthorized`. A valid key whose\nscopes do not authorize a tool gets `403 Forbidden`.\n\nTo inspect key identity and permissions, call `GET /api/v1/system/context` with the API key.\n\n## Client configuration\n\n```json\n{\n \"mcpServers\": {\n \"langbot\": {\n \"url\": \"http://:5300/mcp\",\n \"headers\": { \"X-API-Key\": \"\" }\n }\n }\n}\n```\n\n## Tool surface\n\nSlack quick setup: create a disabled `slack-omni` bot draft first, then use\n`start_slack_setup` with the user's App Configuration access/refresh tokens.\nChoose `socket_mode=true` with a browser-reachable OAuth `redirect_url`, or\nprovide the public HTTPS `webhook_url` ending in `/bots/`.\nPoll `get_slack_setup_status` for the Slack installation authorization URL;\nthe user must authorize installation. On success, apply the returned `config`\nusing the normal bot update flow. Socket Mode also requires a separately\ngenerated `app_token` (`xapp-`, scope `connections:write`). Never log tokens.\n`cancel_slack_setup` removes the temporary session without deleting the Slack\napplication. Sessions expire after 15 minutes and are bound to the initiating\nWorkspace, principal, and placement generation. Restarting a failed setup can\ncreate another Slack application; inspect the returned `app_id` first.\n\nThe tools wrap the LangBot service layer. Current tools (v1):\n\n`create_pipeline` and HTTP `POST /api/v1/pipelines` preserve a supplied `config`\nand `extensions_preferences`; omit `config` to use installed Runner defaults.\nSelecting `ai.runner.id` never grants extension access: an empty `plugins` list\nwith `enable_all_plugins=false` still denies that plugin's Tool, EventListener\nand Command components, while allowing its selected Runner. Editing and copying\na pipeline retain these independent settings.\n\n| Tool | Purpose |\n| --- | --- |\n| `get_system_info` | Version, edition, instance id |\n| `list_bots` / `get_bot` / `create_bot` / `update_bot` / `delete_bot` | Manage messaging-platform bots (secrets redacted on read) |\n| `list_bot_event_route_statuses` | Inspect bot event-route runtime status |\n| `list_processors` / `get_processor` / `create_processor` / `update_processor` / `delete_processor` | Manage the peer Agent, Pipeline and Event processor types |\n| `get_processor_metadata` | Discover installed event-capable Runner components, schemas and supported event patterns. |\n| `list_processor_runs` / `get_processor_run_events` | Read one Agent or plugin processor run history and logs; paginate with `before_id` / `after_sequence`. |\n| `debug_agent` | Execute a synthetic Agent event (`processor_uuid`, `payload`); requires `runtime.operate`. Returns final text and up to 1000 execution events (thinking, text, tool arguments/results). Platform tools use Mock; other configured tools execute normally. Optional `payload.mock`: `errors`/`results` keyed by platform tool name, `unsupported_apis` lists unavailable platform APIs. |\n| `list_pipelines` / `get_pipeline` / `create_pipeline` / `update_pipeline` / `delete_pipeline` | Manage pipelines |\n| `list_llm_models` / `get_llm_model` / `list_embedding_models` / `list_model_providers` | Inspect models & providers |\n| `list_knowledge_bases` / `get_knowledge_base` / `retrieve_knowledge_base` | RAG knowledge bases (incl. semantic search) |\n| `list_mcp_servers` | External MCP servers LangBot connects to (as a client) |\n| `list_skills` / `get_skill` | Installed skills |\n| `list_knowledge_engines` / `get_knowledge_engine_schema` / `list_knowledge_parsers` | Discover RAG configuration |\n| `get_pipeline_extensions` / `update_pipeline_extensions` | Read or completely replace extension bindings; all lists and switches required |\n| `run_pipeline` | One fresh-session turn; requires `runtime.operate`, executes configured models/tools, never auto-retry an unknown outcome |\n| `get_monitoring_records` / `get_monitoring_details` | Bounded Workspace records and existing message/session details |\n| `get_sandbox_diagnostics` | Read status (`resource.view`), sessions/errors (`audit.view`); managed sandbox admission still applies |\n\nMutating tools (`create_*`, `update_*`) take a JSON object matching the same\nshape as the corresponding HTTP API request body. Discover resources with the\n`list_*` / `get_*` tools before mutating; identifiers are UUIDs. Reads require\n`resource.view`; mutations require `resource.manage`. All service calls inherit\nthe immutable Workspace context authenticated at the MCP transport boundary.\nPass `is_default: true` to `create_pipeline` only when the Workspace does not\nalready have a default pipeline.\n\n## How to use\n\nPipeline plugin extension switches authorize only EventListener, Tool and Command\ncomponents. Select the Runner independently in AI settings. An empty extension\nallowlist does not prevent saving, copying or running that Runner. Never enable\nall plugins or add its plugin to the allowlist as a Runner workaround: that also\nauthorizes its other components. The Runner plugin must still be installed and\nenabled in the authenticated Workspace, with an available Plugin Runtime.\n\n1. Get an API key (web UI key, or set `api.global_api_key` in config.yaml).\n2. Point your MCP client at `http://:5300/mcp` with the key header.\n3. Call `get_system_info` to confirm connectivity.\n4. Use `list_*` tools to discover, then `get_*` / `create_*` / `update_*` /\n `delete_*` as needed.\n\n## ChatGPT / Codex subscription providers\n\n`list_model_providers` can return the `openai-codex` requester. Its OAuth\ncredentials are server-only and are not provider API keys. Never ask a user\nto paste ChatGPT access tokens, refresh tokens, or a Codex auth cache into an\nMCP tool or model configuration.\n\nA human connects or disconnects the subscription through **Models → provider\nsettings** in the LangBot web UI. The provider-scoped `/codex/*` authentication\nroutes deliberately require a browser-user session and are not exposed as MCP\ntools or authorized by a LangBot API key. Once connected, models are managed\nand selected through the normal provider/model workflow. A disconnected\nprovider must be reauthorized; do not silently replace it with API-key billing.\n\nSee [ChatGPT / Codex subscription](../../../docs/CODEX_SUBSCRIPTION.md) for setup,\nusage limits, and the personal-account versus shared-service boundary.\n\n## Provider deletion\n\nThe curated MCP surface currently lists providers but has no provider-deletion\ntool. In the web UI, **Edit Provider → Delete** asks for confirmation before\nremoving that provider and all its LLM, embedding, and rerank models. This is\nirreversible; never interpret a request to edit a provider as authorization to\ndelete it.\n\nThe equivalent HTTP operation is\n`DELETE /api/v1/provider/providers/{uuid}?cascade=true`, requiring\n`resource.manage` in the authenticated Workspace. Omitting `cascade` preserves\nthe existing refusal to delete providers that still have models. Cloud-managed\nproviders remain protected. Cascade deletion removes stored Codex authorization\nstate as well; it is not the same operation as disconnecting an account.\n\n## Implementation & maintenance (for LangBot developers)\n\n- Server: `src/langbot/pkg/api/mcp/server.py` (FastMCP). Tools call the service\n layer directly, so the MCP surface stays aligned with the API.\n- Mount: `src/langbot/pkg/api/mcp/mount.py` — an ASGI dispatcher fronting Quart,\n authenticating `/mcp` requests, running the streamable-HTTP session manager.\n- Smoke test: `tests/manual/mcp_smoke.py`.\n\n> When you add, remove, or change an HTTP API endpoint that should be\n> agent-accessible, update the corresponding MCP tool **and** this skill. The\n> MCP tool surface and the API must stay aligned (see `AGENTS.md`).\n\n## Pitfalls\n\n- `/mcp` is the **server** LangBot exposes. The `/api/v1/mcp` routes are the\n **client** side (managing external MCP servers LangBot connects to). Don't\n confuse them.\n- A `401` means the key is wrong, missing, revoked, expired, or (for the global\n key) `api.global_api_key` is empty or the instance is not an OSS singleton.\n- A `403` means the key is valid but lacks the permission required by the tool.\n- The global key is plaintext in config.yaml — only enable it on trusted/internal\n deployments and serve over HTTPS.\n\n## Default message routes\n\n`create_bot` without non-empty `event_bindings` adds two explicit message routes:\nprivate chat (`chat_type eq person`) and group chat (`chat_type eq group` plus\n`is_at_bot eq true`). Both use the Workspace's default Pipeline, or the first\navailable Pipeline. If none exists, first-bot setup creates one with the full\ndefault config. Creating a Pipeline via `create_agent(kind=\"pipeline\")` also uses\nthat full default config. Configure an authorized model before testing replies.\n\nExplicit routes on create are preserved; `update_bot` never automatically\nreplaces manual routes, including an intentionally empty list. The WebUI message\ntemplate uses the same private/group conditions and preserves existing routes\nwhen applied repeatedly. Unmentioned group messages do not enter the Pipeline.\nAn existing broad route can still match first; inspect route order and diagnostics\nbefore changing a member's configuration.\n\n## Event processors\n\nCreate a processor with `kind: \"event_processor\"` and basic information. Without\na component it supports no events. Discover installed components with\n`get_processor_metadata`, then use `update_processor` with `component_ref` and\noptional `parameters`. API callers may also supply these when creating an instance.\nBind an instance by updating the bot's `plugin_processors` array with\n`{\"processor_uuid\": \"\", \"enabled\": true}`. This replaces the full\nsubscription list; preserve bindings you want to keep. Do not add plugin processors\nto `event_bindings`, which remains exclusive Agent/Pipeline routing.\nEach enabled subscription independently receives the installed Runner's declared\nevents. Slow or failed subscribers do not prevent other subscribers or the primary\nroute from executing. Installation alone never activates a handler. Reusing an\ninstance shares its configuration and runtime state. Use a separate instance for\nindependent settings. Optional plugin behavior belongs in the Runner config schema.\n`debug_agent` accepts the complete typed event in `payload.data` for this kind.\nLegacy EventListener plugins remain in the Pipeline lifecycle.\n\n`list_processor_runs` includes `created_at_ms`, `started_at_ms`, and\n`finished_at_ms`: Host lifecycle times in epoch milliseconds. Use the start and finish times for elapsed processing time; select a run and call `get_processor_run_events` for its\nlogs and action results. These times are not internal plugin profiling data.\n\n## Unified execution monitoring\n\nUse `get_inflight_executions` for a bounded current-workspace snapshot of active\nand recently started executions, including `progress_event`. Progress is a\nreported stage, not an estimated completion percentage. The UI uses the shared\n`GET /api/v1/monitoring/in-flight/stream` SSE feed instead of polling per viewer.\n\nUse `get_monitoring_executions` for the execution list and\nits legacy `pipeline_ids` parameter to filter any processor kind (Agent,\nPipeline or event processor); the summary uses the same processor scope. Use\n`get_monitoring_execution_detail` with `source=auto` to resolve a run,\nmessage or event identifier outside the current list page. The detail exposes\n`inputs`, `outputs` (generated content), `deliveries` (recorded platform sends),\n`events`, `llm_calls`, `tool_calls`, `errors`, `related`, and `conversation` in\n`pages`. Follow each section's `has_more` and `next_offset` independently.\nStart with `section=overview` to load only inputs, outputs, deliveries and errors.\nRequest `conversation`, `related`, calls and events separately when needed;\nomitting `section` retains the complete trace response for existing consumers.\nConversation history supplies context; historical messages without explicit\nlinks must not be asserted to belong to the selected execution. Events without\na processor run use `source=event`. All lookups remain Workspace-scoped.\n\nMonitoring record filters, except for `kind=\"sessions\"`, accept `mode` (`all`, `real`, `debug`) and\n`execution_statuses` (normalized execution statuses). These select the owning\nexecution, not the individual model/tool call outcome. Calls without a recorded\nexecution link are excluded when an execution filter is active.\n\nSession lists support `bot_ids`, `pipeline_ids`, `start_time`, `end_time`,\n`user_query`, `is_active`, and pagination. For execution status or real/debug\nfilters, use `get_monitoring_executions`; HTTP `/monitoring/sessions` rejects\nthem. In HTTP `/monitoring/data`, `sessions` and `totalCount.sessions` use only\nbot/pipeline/time filters; execution filters apply to the other records.\n\n### Workspace message language\n\n`get_workspace_message_language` reads the current Workspace's backend language\n(`workspace.view`). `set_workspace_message_language(language)` requires\n`workspace.update`; supported values are `en_US`, `zh_Hans`, `zh_Hant`, `ja_JP`,\n`es_ES`, `ru_RU`, `th_TH`, `vi_VN`. HTTP equivalents are GET/PUT\n`/api/v1/workspaces/current/language` (PUT body: `{\"language\":\"ja_JP\"}`).\nThe API key's Workspace remains authoritative. The setting controls new backend\nmessages and default bot notices, independently of personal UI language.\nPipeline `output.misc` hint fields override defaults; empty values inherit this\nsetting. Existing custom text and historical logs are not rewritten.\n\n### Workspace default LLM model\n\n`get_starred_model` reads the Workspace-wide favorite. `set_starred_model`\nrequires `provider_secret.manage` and replaces the single starred model; pass\n`null` to clear it. The model must belong to the current Workspace.\n`get_default_model` resolves the favorite first, then the Space wizard chat\nrecommendation only when LangBot Models is enabled and the Workspace owner is\nbound to a LangBot Account. It returns a nullable `uuid`.\nNew processor LLM selector defaults use this preference; existing configurations\nare not rewritten. Embedding and rerank models are not eligible.\n\n### Reset a bot session context\n\nUse `reset_session_context(bot_id, session_id)` only when a user asks to start a session afresh. It requires `resource.manage`, preserves monitoring records, and refuses sessions with active tasks. It clears conversation runner state and excludes earlier transcript entries from subsequent model context. Files and long-term memory are not removed.\n\n## Cloud plan platform restrictions\n\nBot creation and enabling use the same Workspace entitlement policy as HTTP. A disabled platform is rejected even through MCP or plugin calls; do not retry using an Omni or legacy alias to bypass the restriction. The Workspace-scoped HTTP adapter catalog omits disabled platforms.\n\n## LLM request diagnostics\n\n`get_monitoring_records(kind=\"llm_calls\")` includes nullable `details` on each call.\nRead `details.timing`, `details.usage`, and `details.error` to distinguish provider\nlatency, retry waits, cached/reasoning tokens, and failures. Timing values are\nmilliseconds since invocation start; `first_event_ms` includes reasoning/control\nframes, while `first_content_ms` measures the first upstream content delta.\n`first_visible_chunk_ms` may contain displayed reasoning, not just answer text.\nOld records have no details. Do not equate total duration with provider TTFT.\nSee `docs/LLM_REQUEST_POLICY.md` for policy precedence and gateway header behavior.\n\n`get_llm_model` also exposes optional `reasoning_capabilities.budget_tokens`\nmin/max bounds. Numeric thinking budgets use `reasoning_config.budget_tokens`\nwith level `enabled`; do not combine budgets with effort levels or native thinking\narguments. Existing Runner reasoning overrides take precedence. Prompt caching\nis opt-in through model/call `extra_args.prompt_caching`; use provider-specific\nmodes documented in `docs/LLM_REQUEST_POLICY.md`. Cache keys are routing hints,\nnot evidence of hits; read the actual usage counters. `preparation_ms` measures\nrequest construction and `preparation_cache_hits` counts retry preparation reuse.\n`transport_setup_ms` measures managed pool acquisition, including first-use TLS\ninitialization; it is not network TTFT. See `docs/LLM_LATENCY_BENCHMARK.md` for\ncontrolled request-layer measurements and their scope.\n","skills/skills/langbot-plugin-dev/SKILL.md":"---\nname: langbot-plugin-dev\ndescription: Develop, debug, and test LangBot plugins. Use when creating new LangBot plugins, fixing plugin bugs, setting up a LangBot test environment, or testing plugins via WebSocket. Covers plugin component architecture (EventListener, Command, Tool), the plugin SDK API (invoke_llm, get_llm_models, send_message, plugin storage), common pitfalls, and automated WebSocket-based testing. Triggers on \"langbot plugin\", \"lbp\", \"GroupChatSummary\", \"plugin debug\", \"langbot test\".\n---\n\n# LangBot Plugin Development & Debugging\n\n## Controlling a running instance via MCP\n\nBeyond writing code, you can **drive a live LangBot instance over MCP** — no raw\nHTTP needed. Two MCP servers exist (both reuse existing API keys; see `AGENTS.md`):\n\n- **LangBot instance** — `http://:5300/mcp` (auth: web-UI `lbk_` key or the\n `api.global_api_key` from `config.yaml`). Manage bots, pipelines, models,\n knowledge bases, and skills. See the **`langbot-mcp-ops`** skill.\n- **LangBot Space marketplace** — `https://space.langbot.app/mcp` (auth: Personal\n Access Token). Search plugins / MCP servers / skills. See the\n **`langbot-space-ops`** skill.\n\n> Any change to an agent-accessible HTTP API endpoint must keep the matching MCP\n> tool and these skills in sync.\n\n## Plugin Architecture\n\nA LangBot plugin consists of:\n\n```\nMyPlugin/\n├── manifest.yaml # Plugin metadata, config schema\n├── main.py # BasePlugin subclass (entry point, shared state)\n├── components/\n│ ├── event_listener/ # Hook pipeline events\n│ │ ├── collector.yaml\n│ │ └── collector.py\n│ ├── commands/ # !command handlers\n│ │ ├── mycommand.yaml\n│ │ └── mycommand.py\n│ └── tools/ # LLM function-call tools\n│ ├── mytool.yaml\n│ └── mytool.py\n```\n\nEach component has a `.yaml` (metadata) and `.py` (implementation).\n\n## README & i18n convention (enforced on the marketplace)\n\nA plugin published to LangBot Space serves a localized README on its detail page.\nThe resolver (`langbot-space` `PluginService.GetPluginREADME`) works like this:\n\n- **Root `README.md` MUST be in English.** It is the default and the fallback —\n when no per-language README matches the viewer's locale, the page serves the\n root `README.md`. A non-English root README makes the English/default view show\n the wrong language.\n- **All other languages live under `readme/README_{lang}.md`** — e.g.\n `readme/README_zh_Hans.md`, `readme/README_ja_JP.md`. The 8 supported locales:\n `en_US, zh_Hans, zh_Hant, ja_JP, th_TH, vi_VN, es_ES, ru_RU`.\n- `manifest.yaml` `metadata.label` / `metadata.description` should carry the same\n 8-locale i18n set (`repository` must be a real, alive URL).\n\n```\nMyPlugin/\n├── manifest.yaml\n├── README.md # English (default + fallback) — REQUIRED, must be English\n└── readme/\n ├── README_zh_Hans.md\n ├── README_zh_Hant.md\n ├── README_ja_JP.md\n ├── README_th_TH.md\n ├── README_vi_VN.md\n ├── README_es_ES.md\n └── README_ru_RU.md\n```\n\n`manifest.yaml` (incl. `repository`) is the source of truth — the marketplace\nsyncs from it, so edit the package and re-publish rather than patching live data.\n\n## Critical SDK Pitfalls\n\n### 1. MessageChain is a RootModel — iterate directly\n\n```python\n# ❌ WRONG — MessageChain has no .components attribute\nfor component in event.message_chain.components:\n\n# ✅ CORRECT — MessageChain is a Pydantic RootModel, iterate directly\nfor component in event.message_chain:\n```\n\n### 2. Message.content must be `list[ContentElement]` or `str`, not a single ContentElement\n\n```python\nfrom langbot_plugin.api.entities.builtin.provider import message as provider_message\n\n# ❌ WRONG — single ContentElement\nMessage(role=\"user\", content=ContentElement.from_text(\"hello\"))\n\n# ✅ CORRECT — list of ContentElement\nMessage(role=\"user\", content=[ContentElement.from_text(\"hello\")])\n\n# ✅ ALSO CORRECT — plain string\nMessage(role=\"user\", content=\"hello\")\n```\n\n### 3. invoke_llm does NOT accept timeout\n\n```python\n# ❌ WRONG\nawait self.invoke_llm(llm_model_uuid=uuid, messages=msgs, timeout=60)\n\n# ✅ CORRECT\nawait self.invoke_llm(llm_model_uuid=uuid, messages=msgs)\n```\n\n### 4. invoke_llm response.content can be str OR list\n\n```python\nresponse = await self.invoke_llm(...)\nif response.content:\n if isinstance(response.content, str):\n return response.content\n elif isinstance(response.content, list):\n parts = [e.text for e in response.content if hasattr(e, \"text\") and e.text]\n return \"\\n\".join(parts)\n```\n\n### 5. get_llm_models() returns UUIDs\n\n```python\n# Returns list[str] of model UUIDs\nmodels = await self.get_llm_models()\nmodel_uuid = models[0] # First available model UUID\n```\n\n**Known bug (v4.9.3):** The host handler may return `list[dict]` instead of `list[str]`. If you hit `TypeError: unhashable type: 'dict'` in `invoke_llm`, the fix is in `LangBot/src/langbot/pkg/plugin/handler.py` — change `'llm_models': llm_models` to `'llm_models': [m['uuid'] for m in llm_models]`.\n\n### 6. invoke_llm parameter is `llm_model_uuid`, NOT `model_uuid`\n\n```python\n# ❌ WRONG — will throw \"got an unexpected keyword argument\"\nawait self.invoke_llm(messages=msgs, model_uuid=uuid)\n\n# ✅ CORRECT\nawait self.invoke_llm(messages=msgs, llm_model_uuid=uuid)\n```\n\n### 7. prevent_default() alone does NOT block LLM response\n\nTo fully prevent the default LLM pipeline from responding when your EventListener handles the message, you must call **both**:\n\n```python\nevent_context.prevent_default() # Block default behavior\nevent_context.prevent_postorder() # Block subsequent plugins/pipeline\n```\n\nUsing only `prevent_default()` still allows the LLM to generate a response.\n\n### 8. get_plugin_storage / set_plugin_storage may throw KeyError: 'owner'\n\nThis is a version mismatch between the SDK and host. Wrap storage calls in try/except:\n\n```python\ntry:\n data = await self.get_plugin_storage(\"my_key\")\nexcept Exception:\n data = None # Fallback gracefully\n```\n\n### 9. Component YAML must have full structure, not just name/description\n\n```yaml\n# ❌ WRONG — will silently fail to register the component\nname: translator\ndescription:\n en_US: 'Does stuff'\n\n# ✅ CORRECT — full component YAML\napiVersion: v1\nkind: EventListener\nmetadata:\n name: translator\n label:\n en_US: Translator\nspec:\nexecution:\n python:\n path: translator.py\n attr: Translator\n```\n\n### 10. BasePlugin import path\n\n```python\n# ❌ WRONG\nfrom langbot_plugin.api.definition.base_plugin import BasePlugin\n\n# ✅ CORRECT\nfrom langbot_plugin.api.definition.plugin import BasePlugin\n```\n\n## Pipeline Events\n\nEvents the EventListener can hook (from most general to most specific):\n\n| Event | When |\n|---|---|\n| `GroupMessageReceived` | **Any** group message arrives (before trigger rules) |\n| `PersonMessageReceived` | **Any** private message arrives |\n| `GroupNormalMessageReceived` | Group message passes trigger rules, going to LLM |\n| `PersonNormalMessageReceived` | Private message going to LLM |\n| `GroupCommandSent` | Group message matched as command |\n| `PersonCommandSent` | Private message matched as command |\n| `NormalMessageResponded` | LLM generated a response |\n| `PromptPreProcessing` | About to build LLM context |\n\n**Key insight:** `*MessageReceived` fires for ALL messages regardless of trigger rules. `*NormalMessageReceived` only fires for messages that match the pipeline's trigger rules (e.g., @bot, prefix, random%). Use `*MessageReceived` for message collection/logging.\n\n## EventContext API\n\n```python\n@self.handler(events.GroupMessageReceived)\nasync def on_msg(event_context: context.EventContext):\n event = event_context.event\n event.launcher_id # Group ID\n event.sender_id # Sender ID\n event.message_chain # MessageChain (iterate directly)\n\n # Reply to the current conversation\n await event_context.reply(MessageChain([Plain(text=\"hello\")]))\n\n # Block default pipeline behavior\n event_context.prevent_default()\n\n # Block subsequent plugins\n event_context.prevent_postorder()\n```\n\n## Setting Up a Test Environment\n\n### Deploy via Docker (GitOps + Portainer)\n\nSee `references/test-env-setup.md` for full deployment steps.\n\nQuick summary:\n1. Create `docker-compose.yaml` in `server-deploy` repo\n2. Deploy via Portainer git repository method\n3. Set up admin account via `/api/v1/user/init` POST\n4. Configure LLM provider and model via API\n5. Copy plugin to `data/plugins/` directory\n\n### WebSocket Testing\n\nLangBot's WebUI chat uses WebSocket. Connect to test message flow:\n\n```\nws://:/api/v1/pipelines//ws/connect?session_type=group\n```\n\n- `session_type=group` for group chat simulation\n- `session_type=person` for private chat (always triggers pipeline)\n\n**Requires Origin header** to pass CORS:\n```javascript\nconst ws = new WebSocket(url, {\n headers: { Origin: 'https://your-langbot-domain' }\n});\n```\n\nSend messages:\n```json\n{\"type\": \"message\", \"message\": [{\"type\": \"Plain\", \"text\": \"hello\"}]}\n```\n\nReceive:\n- `{\"type\": \"connected\", ...}` — connection established\n- `{\"type\": \"user_message\", \"data\": {...}}` — echo of sent message\n- `{\"type\": \"response\", \"data\": {\"content\": \"...\", \"is_final\": true/false}}` — bot reply (streamed)\n\n### Group Trigger Rules\n\nGroup messages only enter the pipeline if trigger rules are met:\n\n```json\n{\n \"group-respond-rules\": {\n \"at\": true, // Respond when @bot\n \"prefix\": [\"ai\"], // Respond to messages starting with \"ai\"\n \"random\": 0.0, // Probability of responding to any message (0.0-1.0)\n \"regexp\": [] // Regex patterns\n }\n}\n```\n\nFor testing, set `random: 1.0` via PUT `/api/v1/pipelines/` to respond to all messages.\n\n**Important:** EventListener hooks like `GroupMessageReceived` fire regardless of trigger rules. Only the LLM processing (`GroupNormalMessageReceived` and beyond) requires trigger rules.\n\n### Plugin Hot-Reload\n\nThere is **no hot-reload**. After changing plugin files:\n\n```bash\ndocker restart \n# Wait ~5 seconds for plugin to re-mount\n```\n\nThe main LangBot container does NOT need restart for plugin changes — only the runtime container.\n\n## API Quick Reference\n\n### Admin Setup\n\n```bash\n# Initialize admin account (first time only)\ncurl -X POST $BASE/api/v1/user/init \\\n -H \"Content-Type: application/json\" \\\n -d '{\"user\":\"admin@test.com\",\"password\":\"test123\"}'\n\n# Login\ncurl -X POST $BASE/api/v1/user/auth \\\n -H \"Content-Type: application/json\" \\\n -d '{\"user\":\"admin@test.com\",\"password\":\"test123\"}'\n# Returns: {\"data\":{\"token\":\"eyJ...\"}}\n```\n\n### Provider & Model Setup\n\n```bash\n# Create provider\ncurl -X POST $BASE/api/v1/provider/providers \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d '{\"name\":\"MyProvider\",\"requester\":\"new-api-chat-completions\",\"base_url\":\"https://api.example.com/v1\",\"api_keys\":[\"sk-xxx\"]}'\n\n# Create LLM model\ncurl -X POST $BASE/api/v1/provider/models/llm \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d '{\"name\":\"gpt-4o-mini\",\"provider_uuid\":\"\",\"abilities\":[\"chat\",\"tool-use\"]}'\n\n# List models\ncurl $BASE/api/v1/provider/models/llm -H \"Authorization: Bearer $TOKEN\"\n```\n\n### Pipeline Config\n\n```bash\n# Get pipeline\ncurl $BASE/api/v1/pipelines -H \"Authorization: Bearer $TOKEN\"\n\n# Update pipeline (e.g., set model, modify trigger rules)\ncurl -X PUT $BASE/api/v1/pipelines/ \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d ''\n```\n\n## Plugin Config Types\n\nSupported `type` values in `manifest.yaml` `spec.config`:\n\n| Type | Description | Value |\n|---|---|---|\n| `string` | Text input | string |\n| `int` / `integer` | Number input | int |\n| `float` | Decimal input | float |\n| `bool` / `boolean` | Toggle | bool |\n| `select` | Dropdown (needs `options`) | string |\n| `prompt-editor` | Multi-line prompt editor | string |\n| `llm-model-selector` | LLM model picker UI | UUID string |\n| `bot-selector` | Bot picker UI | UUID string |\n\nExample — let users choose which model the plugin uses:\n\n```yaml\nspec:\n config:\n - name: model\n type: llm-model-selector\n label:\n en_US: 'LLM Model'\n zh_Hans: 'LLM 模型'\n description:\n en_US: 'Select the LLM model. Falls back to first available if not set.'\n zh_Hans: '选择 LLM 模型。未设置时使用第一个可用模型。'\n required: false\n```\n\nRead config in plugin code:\n\n```python\nmodel_uuid = self.get_config().get(\"model\")\n```\n\n## Container Restart Timing\n\nAfter plugin file changes, **only the runtime container needs restart**:\n\n```bash\ndocker restart langbot-test-runtime\n# Wait ~15 seconds before testing\n```\n\n**When to restart both (runtime first, then host):**\n- Added/removed Command or Tool components (host caches component lists)\n- Changed `manifest.yaml` structure\n\n```bash\ndocker restart langbot-test-runtime\nsleep 8\ndocker restart langbot-test\nsleep 8\n```\n\n**⚠️ Do NOT restart both simultaneously** — the host may connect before plugins are mounted, causing 502 errors or missing plugin registrations.\n\n## Debugging Checklist\n\nWhen a plugin doesn't work:\n\n1. **Check runtime logs**: `docker logs ` — look for mount/init errors\n2. **Check host logs**: `docker logs ` — look for pipeline processing errors\n3. **Verify plugin loaded**: `GET /api/v1/plugins` — should list your plugin\n4. **Test person mode first**: `session_type=person` always triggers pipeline, isolating trigger rule issues\n5. **Check trigger rules**: Group mode requires @bot, prefix match, or random% to enter pipeline\n6. **Verify model configured**: Pipeline's `config.ai.runner_config[config.ai.runner.id].model.primary` must point to a valid model UUID with working API keys\n\n## Publishing Plugins\n\nAfter testing, publish via `lbp publish`:\n\n```bash\ncd /path/to/MyPlugin\nlbp publish\n```\n\nThis builds `.lbpkg` and uploads to Space marketplace as a draft. Then go to https://space.langbot.app/market to upload screenshots and submit for review.\n\n**Prerequisite:** Must be logged in via `lbp login --token lbpat_xxx` (PAT from Space profile page).\n\n## Reference: EventListener-Only Plugin Pattern\n\nFor plugins that react to messages without commands or tools (e.g., auto-summarize URLs, collect messages, translate):\n\n```\nMyPlugin/\n├── manifest.yaml # Only EventListener in spec.components\n├── main.py # BasePlugin with shared logic (fetch, LLM calls)\n├── components/\n│ └── event_listener/\n│ ├── detector.yaml\n│ └── detector.py\n└── requirements.txt\n```\n\n**manifest.yaml** — only declare EventListener:\n```yaml\nspec:\n components:\n EventListener:\n fromDirs:\n - path: components/event_listener/\n```\n\n**detector.py** — hook `*MessageReceived`, extract text, process, reply:\n```python\n@self.handler(events.PersonMessageReceived)\nasync def on_msg(event_context: context.EventContext):\n event = event_context.event\n text_parts = []\n for component in event.message_chain:\n if isinstance(component, platform_message.Plain):\n text_parts.append(component.text)\n text = \"\".join(text_parts).strip()\n \n if should_handle(text):\n event_context.prevent_default()\n event_context.prevent_postorder()\n result = await self.plugin.process(text)\n await event_context.reply(platform_message.MessageChain([\n platform_message.Plain(text=result)\n ]))\n```\n\n**Key:** Access shared plugin logic via `self.plugin` (the BasePlugin instance).\n","skills/skills/langbot-plugin-dev/references/test-env-setup.md":"# Test Environment Setup\n\n## Docker Compose (GitOps)\n\nCreate in `server-deploy` repo under `servers//langbot-test/docker-compose.yaml`:\n\n```yaml\nversion: \"3\"\nservices:\n langbot_plugin_runtime:\n image: rockchin/langbot:latest\n container_name: langbot-test-runtime\n volumes:\n - /opt/docker-data/langbot-test/data/plugins:/app/data/plugins\n ports:\n - \"5411:5401\"\n restart: on-failure\n environment:\n - TZ=Asia/Shanghai\n command: [\"uv\", \"run\", \"--no-sync\", \"-m\", \"langbot_plugin.cli.__init__\", \"rt\"]\n networks:\n - langbot_test_network\n\n langbot:\n image: rockchin/langbot:latest\n container_name: langbot-test\n volumes:\n - /opt/docker-data/langbot-test/data:/app/data\n ports:\n - \"5310:5300\"\n restart: on-failure\n depends_on:\n - langbot_plugin_runtime\n environment:\n - TZ=Asia/Shanghai\n networks:\n - langbot_test_network\n\nnetworks:\n langbot_test_network:\n driver: bridge\n```\n\n## Post-Deploy Configuration\n\nAfter first start, LangBot auto-generates `data/config.yaml`. You need to update `plugin.runtime_ws_url` to match the runtime container name:\n\n```bash\n# On the host, edit config\nsed -i 's|ws://localhost:5400/control/ws|ws://langbot-test-runtime:5400/control/ws|' \\\n /opt/docker-data/langbot-test/data/config.yaml\ndocker restart langbot-test\n```\n\n## Installing a Plugin\n\nCopy plugin directory to `data/plugins/` on the host:\n\n```bash\nscp -r MyPlugin/ user@host:/opt/docker-data/langbot-test/data/plugins/MyPlugin/\ndocker restart langbot-test-runtime # Runtime picks up new plugins on restart\n```\n\n## Caddy Reverse Proxy (Optional)\n\nIf testing externally, add to Caddyfile on the same host:\n\n```\nlangbot-test.example.com {\n reverse_proxy langbot-test:5300\n}\n```\n\nThen reload: `docker exec caddy caddy reload --config /etc/caddy/Caddyfile`\n\nThe WebSocket endpoint works through Caddy without special config.\n\n## WebSocket Test Script (Node.js)\n\n```javascript\nconst WebSocket = require('ws');\n\nconst PIPELINE_UUID = '';\nconst BASE = 'wss://langbot-test.example.com';\nconst URL = `${BASE}/api/v1/pipelines/${PIPELINE_UUID}/ws/connect?session_type=group`;\n\nconst ws = new WebSocket(URL, {\n headers: { Origin: BASE }\n});\n\nconst send = (text) => {\n ws.send(JSON.stringify({\n type: 'message',\n message: [{ type: 'Plain', text }]\n }));\n console.log('[SENT]', text);\n};\n\nws.on('message', (data) => {\n const msg = JSON.parse(data.toString());\n if (msg.type === 'connected') {\n console.log('Connected!');\n // Send test messages\n send('Message 1');\n setTimeout(() => send('Message 2'), 500);\n setTimeout(() => send('!summary'), 2000);\n } else if (msg.type === 'response' && msg.data?.is_final) {\n console.log('[BOT]', msg.data.content);\n }\n});\n\nws.on('error', (e) => console.error('Error:', e.message));\nsetTimeout(() => { ws.close(); process.exit(); }, 60000);\n```\n\nRequires: `npm install ws`\n","skills/skills/langbot-skills-maintenance/SKILL.md":"---\nname: langbot-skills-maintenance\ndescription: Maintain the langbot-skills repository with low duplication. Use when adding, editing, or auditing LangBot skills, references, cases, troubleshooting entries, indexes, or periodic entropy-control checks for this skills repository.\n---\n\n# LangBot Skills Maintenance\n\nUse this skill before changing reusable assets in this repository.\n\n## Workflow\n\n1. Read `AGENTS.md`, `skills/.env`, and the relevant existing skill files.\n2. Classify the change:\n - `SKILL.md` for routing and concise operating rules.\n - `references/*.md` for canonical detailed workflows.\n - `cases/*.yaml` for executable test-plan skeletons.\n - `suites/*.yaml` for reusable groups of case ids.\n - `fixtures/fixtures.json` for deterministic fixture readiness metadata.\n - `reports/evidence//automation-result.json` as automation output and `reports/evidence//result.json` as final judgment output; neither is a catalog asset to commit.\n - `troubleshooting/*.yaml` for one reusable failure mode.\n3. Search existing assets before adding new files:\n - `rg \"\" skills`\n - `bin/lbs case list`\n - `bin/lbs suite list`\n - `bin/lbs fixture list`\n4. Put detail in one canonical place and link to it from cases or routing bullets.\n5. Run the checks in `AGENTS.md` after edits.\n\n## Entropy Rules\n\n- Prefer extending an existing reference or troubleshooting entry when the root cause is the same.\n- Keep cases short: setup, action, evidence, pass/fail checks. Do not paste long prompts or debug transcripts when a reference exists.\n- Put machine-checkable inputs in `env`, `automation_env`, or fixtures; put operator-confirmed assumptions in `preconditions` so `test plan` can surface `manual_check`.\n- Keep suites short: title, intent, tags, and ordered case ids. Do not duplicate case steps inside a suite.\n- Keep fixture manifests factual: id, title, path, kind, and related case ids. Do not encode environment-specific absolute paths.\n- Keep troubleshooting entries narrow: symptoms, patterns, likely causes, fixes, related assets.\n- Do not hardcode local ports, browser profile paths, secrets, tokens, or provider keys.\n- Use `bin/lbs index --check` to verify the committed index is current without writing it; run `bin/lbs index` when the index needs regeneration.\n\nFor periodic repository audits, read `references/curation-workflow.md`.\n","skills/skills/langbot-skills-maintenance/references/curation-workflow.md":"# Curation Workflow\n\nUse this checklist when the repository starts accumulating repeated cases, copied steps, or overlapping troubleshooting entries.\n\n## Audit Pass\n\n1. Inspect the current surface:\n - `bin/lbs case list`\n - `bin/lbs case list --json --priority p0 --automation`\n - `bin/lbs case list --ready`\n - `bin/lbs case list --machine-ready`\n - `bin/lbs suite list`\n - `bin/lbs fixture list`\n - `rg \"sandbox|provider|pipeline|plugin|knowledge|mcp\" skills`\n - `rg \"If .* fails|Known Pitfalls|Debug Chat|/api/v1\" skills`\n2. Group nearby assets by intent, not by file path:\n - user-facing scenario\n - backend or provider dependency\n - failure signature\n - pass/fail evidence\n3. Pick one canonical owner:\n - stable procedures belong in `references/`\n - deterministic files and packages belong in `fixtures/` plus `fixtures/fixtures.json`\n - repeated failure signatures belong in `troubleshooting/`\n - runnable QA paths belong in `cases/`\n - reusable groups of QA paths belong in `suites/`\n - skill entry points belong in `SKILL.md`\n\n## Merge Or Split\n\nMerge when two files share the same trigger, root cause, and fix. Keep the stronger id and move missing patterns into it.\n\nSplit when a file mixes unrelated failure modes or requires different fixes. Each troubleshooting id should map to one diagnosis path.\n\nMove repeated step lists out of cases and into a reference when more than one case would need the same prompt, UI path, or log interpretation.\n\nAdd or update a suite when developers repeatedly run the same ordered group of cases. Do not copy case steps into suites; use `bin/lbs suite plan ` to expand the group.\nUse `bin/lbs suite start ` and `bin/lbs suite report --evidence-dir ` when validating that a suite is operational end to end.\n\nAdd or update `fixtures/fixtures.json` when a case depends on a deterministic file, plugin package, or local test server. The manifest should use repo-relative paths under the owning skill and should not contain machine-local absolute paths.\n\nWhen adding Debug Chat Playwright automation, reuse `scripts/e2e/lib/debug-chat.mjs` for navigation, prompt send, response leaf matching, and known failure classification. Keep case-specific prompts and expected sentinels in case YAML automation fields when possible.\n\n## Case Review\n\nFor every changed case:\n\n1. Ensure `steps` describe what to execute, not every command in the underlying implementation.\n2. Ensure `checks` contain observable UI, log, network, or filesystem evidence.\n3. Ensure `diagnostics` are fallback investigation hints, not pass criteria.\n4. Ensure `priority`, `risk`, `ci_eligible`, and `evidence_required` match the actual repeatability and evidence burden.\n5. Put must-have env vars in `env` / `automation_env`; put one-of choices such as URL-or-name in `env_any` / `automation_env_any`.\n6. Ensure linked `skills` and `troubleshooting` ids exist.\n7. Run:\n\n ```bash\n bin/lbs validate\n bin/lbs index --check\n bin/lbs index\n bin/lbs test plan \n ```\n\n## Final Gate\n\nBefore handing off:\n\n- `git diff --stat` should show a focused change set.\n- `skills.index.json` should be regenerated only by `bin/lbs index`.\n- No new asset should contain local credentials, OAuth tokens, API keys, or copied localStorage values.\n- The final note should say which checks ran and which cases or troubleshooting ids changed.\n","skills/skills/langbot-space-ops/SKILL.md":"---\nname: langbot-space-ops\ndescription: Browse and search the LangBot Space marketplaces (plugins, MCP servers, skills) through the Space MCP server. Use when an AI agent needs to discover LangBot extensions on space.langbot.app over MCP. Covers the /mcp endpoint, Personal Access Token (PAT) auth, the tool surface, and client configuration. Triggers on \"langbot space mcp\", \"search langbot plugins\", \"langbot marketplace mcp\", \"space.langbot.app mcp\".\n---\n\n# LangBot Space MCP Operations\n\nLangBot Space (space.langbot.app) exposes an **MCP server** so user-facing AI\nagents can browse and search the marketplaces (plugins, MCP servers, skills) and\nrank live models for automated setup.\n\n## Endpoint\n\n```\nhttps://space.langbot.app/mcp\n```\n\nTransport: **streamable HTTP** (stateless, JSON responses). For a self-hosted\nSpace instance: `http://:8383/mcp`.\n\n## Authentication\n\nReuses the existing **Personal Access Token (PAT)** — the same token the `lbp`\nCLI uses. Create one in your Space account (Profile → Personal Access Tokens),\nthen send it as a Bearer token:\n\n```\nAuthorization: Bearer \n```\n\nRequests without a valid PAT get `401 Unauthorized`.\n\n## Client configuration\n\n```json\n{\n \"mcpServers\": {\n \"langbot-space\": {\n \"url\": \"https://space.langbot.app/mcp\",\n \"headers\": { \"Authorization\": \"Bearer \" }\n }\n }\n}\n```\n\n## Tool surface\n\n| Tool | Purpose |\n| --- | --- |\n| `list_plugins` / `search_plugins` / `get_plugin` | Plugin marketplace |\n| `list_mcp_servers` / `search_mcp_servers` / `get_mcp_server` | MCP-server marketplace |\n| `list_skills` / `search_skills` / `get_skill` | Skill marketplace |\n| `select_models` | Live best-first model list for setup wizards; optional `category` filter |\n\n`list_*` and `search_*` are paged (`page`, `page_size`). `get_*` takes\n`author` + `name`. The tool surface mirrors the REST endpoints under\n`/api/v1/marketplace/*`; `select_models` mirrors `/api/v1/models/selection`.\nAll tools are read-only.\n\n## How to use\n\n1. Create a PAT in your Space account settings.\n2. Point your MCP client at `https://space.langbot.app/mcp` with the Bearer PAT.\n3. Use `search_plugins` / `search_mcp_servers` / `search_skills` to find items,\n then `get_*` for details (e.g. to obtain author/name for installation in\n LangBot itself).\n4. For automatic local-agent setup, call `select_models` (optionally with\n `category`) and choose the first compatible item. Ordering is latest probe\n state (available, unprobed, unavailable), then Space recommendation. Each\n item includes `availability.up`, `last_probed_at`, latency, and HTTP status.\n\n## Runner usage recommendations\n\nUse `search_plugins` with `runner_usage: \"agent\"` for Agent, pipeline, and\nsetup-wizard recommendations, or `runner_usage: \"event\"` for event processors.\nThe component kind remains `Runner`. Only these two exact values are accepted;\nomit the optional field to preserve unfiltered browsing.\n\n```json\n{\"query\":\"\", \"runner_usage\":\"agent\", \"page\":1, \"page_size\":100}\n```\n\nPlugin results include `latest_version` and `runner_usages: string[]`, the\nexplicit union of usages in that latest installable version. Only recommend a\nplugin when this array explicitly contains the target usage. Missing, empty,\nmalformed, or unknown usages must never mean agent-compatible. Event-only\nplugins must never enter Agent recommendations. Empty filtered results are\nvalid while legacy packages await corrected releases; never remove the filter\nto fill a recommendation list.\n\nREST callers use `runner_usage` on both\n`POST /api/v1/marketplace/extensions/search` and the compatibility\n`POST /api/v1/marketplace/plugins/search`; preserve it during fallback. Add\n`\"type_filter\":\"plugin\", \"component_filter\":\"Runner\"` on the unified endpoint.\nUsage is ANDed with other filters before pagination and `total`; MCP/Skill items\ndo not match. Invalid REST values return HTTP 400.\n\nOpen the same filter in the webpage:\n`https://space.langbot.app/market?type=plugin&component=Runner&runner_usage=agent`\n(or `runner_usage=event`). Switch All / Agent / Event in the Runner usage row.\n\n## Implementation & maintenance (for Space developers)\n\n- Server: `internal/controller/mcp/server.go` (official Go MCP SDK\n `github.com/modelcontextprotocol/go-sdk`). Tools call the service layer\n (`PluginService`, `MCPService`, `SkillService`, `ModelStatusService`) directly.\n- Mount: `internal/controller/api.go` at `/mcp` and `/mcp/*any`.\n- Auth: PAT via `AccountService.ValidatePersonalAccessToken`.\n- Docs: `docs/MCP_SERVER.md`.\n\n> When you add, remove, or change a marketplace API endpoint that should be\n> agent-accessible, update the corresponding MCP tool **and** this skill. The\n> MCP tool surface and the API must stay aligned (see `AGENTS.md`).\n\n## Pitfalls\n\n- The PAT prefix is `lbpat_` (Space), distinct from LangBot's `lbk_` API keys.\n- This server is read/browse only; it does not publish or modify marketplace\n items. Use the web UI or REST API (with appropriate auth) for that.\n","skills/skills/langbot-testing/SKILL.md":"","skills/skills/langbot-testing/cases/acp-agent-runner-debug-chat.yaml":"id: acp-agent-runner-debug-chat\ntitle: \"ACP Runner can answer through Debug Chat using real remote Claude\"\nmode: agent-browser\narea: pipeline\ntype: regression\npriority: p2\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - acp\n - claude\n - external-runner\n - pipeline\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nenv_any:\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/pipeline-debug-chat.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\n - LANGBOT_ACP_RUNNER_PIPELINE_URL\n - LANGBOT_ACP_RUNNER_PIPELINE_NAME\n - LANGBOT_E2E_PROMPT\n - LANGBOT_E2E_EXPECTED_TEXT\n - LANGBOT_E2E_EXPECTED_RUNNER_ID\n - LANGBOT_E2E_RESPONSE_TIMEOUT_MS\nautomation_pipeline_url_env: LANGBOT_ACP_RUNNER_PIPELINE_URL\nautomation_pipeline_name_env: LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation_expected_runner_id: \"plugin:langbot-team/ACPRunner/default\"\nautomation_prompt: \"Do not launch any background agent, subagent, task, or worker. In this current ACP session, directly call the MCP tool named langbot_get_current_event exactly once and wait for its result. After it returns, reply exactly ACP_RUNNER_E2E_OK with no other text.\"\nautomation_expected_text: \"ACP_RUNNER_E2E_OK\"\nautomation_response_timeout_ms: \"300000\"\nsetup_automation:\n - \"node:scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env\"\nsetup_provides_env:\n - LANGBOT_ACP_RUNNER_PIPELINE_URL\n - LANGBOT_ACP_RUNNER_PIPELINE_NAME\npreconditions:\n - \"The remote machine has a working Claude Code login and can run npx -y @agentclientprotocol/claude-agent-acp.\"\n - \"LangBot can non-interactively SSH to the remote machine; the runner opens the MCP reverse tunnel automatically.\"\nsteps:\n - \"Open LANGBOT_FRONTEND_URL.\"\n - \"Open the ACP Runner QA pipeline.\"\n - \"Confirm the pipeline AI runner is plugin:langbot-team/ACPRunner/default.\"\n - \"Open Debug Chat.\"\n - \"Ask the real remote Claude ACP agent to call langbot_get_current_event and return ACP_RUNNER_E2E_OK exactly.\"\nchecks:\n - \"UI: Debug Chat shows the user prompt.\"\n - \"UI: Debug Chat shows a Bot response containing ACP_RUNNER_E2E_OK.\"\n - \"Logs: Backend logs include Processing request from person_websocket and Streaming completed for this run.\"\n - \"Logs: No acp runner request error appears for this run.\"\n - \"Console: No unexpected frontend errors appear during Debug Chat.\"\nevidence_required:\n - ui\n - console\n - backend_log\ndiagnostics:\n - \"Use scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env to create/update the pipeline.\"\n - \"For remote Claude on 101, verify ssh yhh@101.34.71.12 can run without password prompts; no separate ssh -R process is required.\"\nsuccess_patterns:\n - \"ACP_RUNNER_E2E_OK\"\n - \"Processing request from person_websocket\"\n - \"Streaming completed\"\nfailure_patterns:\n - \"acp.command_not_found\"\n - \"acp.process_exited\"\n - \"Agent runner plugin:langbot-team/ACPRunner/default execution failed\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - proxy-env-mismatch\n","skills/skills/langbot-testing/cases/agent-run-ledger-audit.yaml":"id: agent-run-ledger-audit\ntitle: \"Persisted Runner run ledger passes end-to-end invariants\"\nmode: probe\narea: agent\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - ledger\n - audit\n - tools\nskills:\n - langbot-testing\nautomation: scripts/e2e/agent-run-ledger-audit.mjs\nsteps:\n - \"Set LANGBOT_AGENT_RUN_ID to audit a specific run, or leave it unset to audit the latest persisted Runner run.\"\n - \"For an external runner's own CLI tools, set LANGBOT_AGENT_TOOL_AUTHORIZATION_MODE=runner-native; keep the default strict mode for Host tool calls.\"\n - \"Read the active LangBot database configuration and inspect the selected run and its ordered events.\"\n - \"Verify completed terminal state, run.completed, paired tool.call.started/completed events, stable tool names, and monotonic ordering.\"\n - \"Compare called tools with the authorization snapshot and validate each advertised tool has owner/source, description, and parameter schema.\"\n - \"Reject malformed persisted event JSON, unrecovered invalid tool arguments, timeout, forbidden, permission-denied, or unauthorized signals.\"\n - \"Record a malformed model tool-argument payload as a recovery warning only when a later tool call succeeds and the run completes.\"\nchecks:\n - \"ledger-audit.json status is pass.\"\n - \"metrics has equal tool_call_started and tool_call_completed counts.\"\n - \"failures is empty and no unauthorized tool call is present.\"\nevidence_required:\n - filesystem\n - metrics\n - api_diagnostic\ndiagnostics:\n - \"Run this immediately after a complex UI Agent task so latest-run selection cannot drift to unrelated traffic.\"\n - \"Use LANGBOT_AGENT_RUN_ID when multiple operators share the test instance.\"\n - \"runner-native mode still checks tool pairing, stable names, ordering, terminal state, and error signals; the external runner's permission system owns its local tools.\"\nsuccess_patterns:\n - \"Agent run ledger audit passed\"\nfailure_patterns:\n - \"tool_call_pairing\"\n - \"forbidden_error_signals\"\n - \"unauthorized_tool_calls\"\ntroubleshooting:\n - box-runtime-silent-action-timeout\n","skills/skills/langbot-testing/cases/agent-runner-async-db-readiness.yaml":"id: agent-runner-async-db-readiness\ntitle: \"Runner async DB readiness probe\"\nmode: probe\narea: release\ntype: smoke\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - async-db\n - aiosqlite\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-async-db-readiness.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-async-db-readiness --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation checks whether a direct aiosqlite in-memory connection can create a table within the readiness timeout.\"\nchecks:\n - \"automation-result.json status is pass or env_issue.\"\n - \"pass means async SQLite tests are worth running.\"\n - \"env_issue means async SQLite-dependent Host pytest probes should be classified as environment-limited until fixed.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"If this probe returns env_issue, run agent-runner-ledger-invariants for fast ledger coverage and skip async ledger pytest as a release blocker in this environment.\"\nsuccess_patterns:\n - \"AIOSQLITE_READY\"\nfailure_patterns:\n - \"aiosqlite readiness timed out\"\ntroubleshooting:\n - aiosqlite-connect-hangs\n","skills/skills/langbot-testing/cases/agent-runner-behavior-matrix.yaml":"id: agent-runner-behavior-matrix\ntitle: \"Runner deterministic behavior matrix probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - deterministic-runner\n - protocol\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-behavior-matrix.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-behavior-matrix --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation reads fixtures/agent-runner/qa-runner-behaviors.json and validates each result sequence through the Host AgentResultNormalizer.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains QA_RUNNER_BEHAVIOR_MATRIX_OK.\"\n - \"The matrix covers ok, stream_ok, empty_output, malformed_result, and controlled_failure.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"fail means the deterministic behavior fixture and Host result normalization disagree.\"\nsuccess_patterns:\n - \"QA_RUNNER_BEHAVIOR_MATRIX_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"RunnerProtocolError\"\n - \"behavior matrix exited\"\n","skills/skills/langbot-testing/cases/agent-runner-fixture-contract.yaml":"id: agent-runner-fixture-contract\ntitle: \"QA Runner fixture contract probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - fixture\n - deterministic-runner\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-fixture-contract.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-fixture-contract --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation imports the QA Runner fixture source and executes normal, streaming, and controlled-failure paths with SDK entities.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains QA_RUNNER_FIXTURE_CONTRACT_OK.\"\n - \"Normal input returns QA_RUNNER_OK:.\"\n - \"Streaming input emits message.delta chunks and completes.\"\n - \"Failure input returns QA_RUNNER_CONTROLLED_FAILURE.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This validates the deterministic fixture source contract. It does not prove the plugin package is installed in a live LangBot instance.\"\nsuccess_patterns:\n - \"QA_RUNNER_FIXTURE_CONTRACT_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"fixture contract exited\"\n","skills/skills/langbot-testing/cases/agent-runner-health-visibility.yaml":"id: agent-runner-health-visibility\ntitle: \"Agent configuration shows whether its selected runner is usable\"\nmode: agent-browser\narea: agent\ntype: feature\npriority: p0\nrisk: medium\nci_eligible: false\ntags:\n - agent-runner\n - agent\n - productization\n - release-gate\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\nautomation: scripts/e2e/agent-runner-health-visibility.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\npreconditions:\n - \"The plugin runtime is enabled and connected, with at least one Runner registered.\"\n - \"The target is a local test instance where a temporary Agent may be created and deleted.\"\nsteps:\n - \"Read the live plugin runtime status and select a registered runner from Agent metadata.\"\n - \"Create a temporary Agent using that runner and open its Runner settings in the WebUI.\"\n - \"Confirm the UI reports that the runner is ready.\"\n - \"Replace the temporary Agent binding with a syntactically valid but unregistered runner id.\"\n - \"Reload Runner settings and confirm the UI reports that the selected runner is unavailable.\"\nchecks:\n - \"UI: A registered runner and connected plugin runtime produce a visible ready status.\"\n - \"UI: A stale runner binding produces a visible unavailable status with recovery actions.\"\n - \"API: Plugin runtime status is connected and Agent metadata contains the registered runner.\"\n - \"Console: No unexpected frontend errors appear during the flow.\"\n - \"Cleanup: The temporary Agent is deleted after evidence is collected.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - api_diagnostic\ndiagnostics:\n - \"A disconnected plugin runtime is env_issue, not a passing unavailable-runner product check.\"\n - \"A ready status inferred only from API metadata without opening the Runner UI is not a pass.\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - proxy-env-mismatch\n","skills/skills/langbot-testing/cases/agent-runner-ledger-concurrency.yaml":"id: agent-runner-ledger-concurrency\ntitle: \"Runner run ledger concurrency and auth pytest probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - pytest\n - run-ledger\n - concurrency\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-concurrency.mjs\npreconditions:\n - \"This Host pytest probe can be slow in the current multi-repo dev environment; keep it in the release gate, but do not treat a timeout as a browser E2E failure without checking pytest logs.\"\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-concurrency --dry-run` first; remove `--dry-run` only after `agent-runner-async-db-readiness` is pass.\"\n - \"Automation resolves LANGBOT_REPO, defaulting to ../LangBot when the env var is unset.\"\n - \"Automation runs selected high-value tests from test_run_ledger_store.py and test_run_ledger_api_auth.py.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"pytest exit status is 0 for selected run ledger claim, lease, status, token, ownership, and active-claim tests.\"\n - \"pytest-stdout.log and pytest-stderr.log are written under LBS_EVIDENCE_DIR.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"env_issue means LANGBOT_REPO/default ../LangBot did not resolve, rtk/uv was unavailable, or the expected test files are missing.\"\n - \"fail means one of the selected LangBot run ledger pytest targets failed or timed out.\"\nsuccess_patterns:\n - \"pytest passed\"\nfailure_patterns:\n - \"pytest exited with status\"\n - \"pytest timed out\"\n - \"Failed to start pytest command\"\ntroubleshooting:\n - aiosqlite-connect-hangs\n","skills/skills/langbot-testing/cases/agent-runner-ledger-contention.yaml":"id: agent-runner-ledger-contention\ntitle: \"Runner ledger SQLite contention probe\"\nmode: probe\narea: release\ntype: regression\npriority: p1\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - stress\n - ledger\n - concurrency\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-contention.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-contention --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation creates 120 queued runs in a file-backed SQLite database and uses eight worker threads to claim runs under write contention.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_CONTENTION_OK.\"\n - \"Every run is claimed once, reaches completed status, and has dispatch_attempts = 1.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe catches obvious exactly-once claim regressions under local SQLite contention; it does not replace async Host pytest or PostgreSQL concurrency checks.\"\nsuccess_patterns:\n - \"LEDGER_CONTENTION_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger contention exited\"\n","skills/skills/langbot-testing/cases/agent-runner-ledger-invariants.yaml":"id: agent-runner-ledger-invariants\ntitle: \"Runner ledger schema and status invariants probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - ledger\n - invariant\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-invariants.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-invariants --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation resolves LANGBOT_REPO, defaulting to the parent LangBot checkout, and imports the sibling SDK from LANGBOT_PLUGIN_SDK_REPO or ../../langbot-plugin-sdk/src.\"\n - \"Automation checks run status sets, terminal status validation, ledger table/index DDL, and a minimal synchronous insert/read path.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_INVARIANTS_OK.\"\n - \"The probe does not require aiosqlite or browser UI.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"env_issue means LANGBOT_REPO/default ../LangBot did not resolve or Python dependencies are unavailable.\"\n - \"fail means a ledger schema/status invariant changed and the release gate needs review.\"\nsuccess_patterns:\n - \"LEDGER_INVARIANTS_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger invariant probe exited\"\n","skills/skills/langbot-testing/cases/agent-runner-ledger-stress.yaml":"id: agent-runner-ledger-stress\ntitle: \"Runner ledger lightweight stress probe\"\nmode: probe\narea: release\ntype: regression\npriority: p1\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - stress\n - ledger\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-stress.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-stress --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation creates 100 queued runs in synchronous SQLite and simulates five runtimes claiming them in priority order.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_STRESS_OK.\"\n - \"Every run is claimed once and reaches a terminal completed status.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe is a fast deterministic stress baseline; it does not replace PostgreSQL/async concurrency tests.\"\nsuccess_patterns:\n - \"LEDGER_STRESS_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger stress exited\"\n","skills/skills/langbot-testing/cases/agent-runner-live-install.yaml":"id: agent-runner-live-install\ntitle: \"QA Runner package installs and registers in LangBot\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - plugin\n - local-install\n - fixture\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_BACKEND_URL\n - LANGBOT_REPO\n - LANGBOT_E2E_LOGIN_USER\nautomation: scripts/e2e/install-qa-plugin-smoke.mjs\nautomation_plugin_package: \"skills/langbot-testing/fixtures/plugins/qa-agent-runner/dist/qa-agent-runner-0.1.0.lbpkg\"\nautomation_expected_plugin_id: \"qa/agent-runner\"\nautomation_expected_tool: \"\"\nautomation_expected_runner_id: \"plugin:qa/agent-runner/default\"\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-live-install --dry-run` first; remove `--dry-run` only after readiness points at a local test LangBot instance.\"\n - \"Automation authenticates the local test user, uploads the QA Runner .lbpkg package, waits for the install task, and reads pipeline metadata.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"/api/v1/plugins lists qa/agent-runner after install.\"\n - \"/api/v1/pipelines/_/metadata lists plugin:qa/agent-runner/default as an available runner.\"\nevidence_required:\n - api_diagnostic\n - filesystem\ndiagnostics:\n - \"This proves the deterministic package installs and registers a runner. It does not prove Debug Chat execution; use a later browser case for that.\"\n - \"If installation fails during dependencies, inspect plugin runtime logs and plugin-dependency-install-offline.\"\nsuccess_patterns:\n - \"qa/agent-runner is installed.\"\nfailure_patterns:\n - \"Plugin install task did not complete successfully\"\n - \"plugin:qa/agent-runner/default is not listed\"\ntroubleshooting:\n - plugin-runtime-timeout\n - plugin-dependency-install-offline\n","skills/skills/langbot-testing/cases/agent-runner-qa-debug-chat.yaml":"id: agent-runner-qa-debug-chat\ntitle: \"QA Runner returns deterministic output through Debug Chat\"\nmode: agent-browser\narea: pipeline\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - pipeline\n - debug-chat\n - fixture\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/pipeline-debug-chat.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation_pipeline_url_env: LANGBOT_QA_RUNNER_PIPELINE_URL\nautomation_pipeline_name_env: LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation_expected_runner_id: \"plugin:qa/agent-runner/default\"\nautomation_prompt: \"hello-live\"\nautomation_expected_text: \"QA_RUNNER_OK:hello-live\"\nautomation_response_timeout_ms: \"120000\"\nautomation_debug_chat_response_p95_ms: \"120000\"\nautomation_reset_debug_chat: \"1\"\nsetup_automation:\n - \"case:agent-runner-live-install\"\n - \"node:scripts/e2e/ensure-qa-agent-runner-pipeline.mjs --write-env\"\nsetup_provides_env:\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nsteps:\n - \"Open LANGBOT_FRONTEND_URL.\"\n - \"Open the pipeline from LANGBOT_QA_RUNNER_PIPELINE_URL or LANGBOT_QA_RUNNER_PIPELINE_NAME.\"\n - \"Confirm the pipeline AI runner is plugin:qa/agent-runner/default.\"\n - \"Open Debug Chat.\"\n - \"Send: hello-live.\"\nchecks:\n - \"UI: The user message appears in Debug Chat.\"\n - \"UI: A Bot message appears and contains QA_RUNNER_OK:hello-live.\"\n - \"API diagnostic: pipeline config uses plugin:qa/agent-runner/default.\"\n - \"Console: No unexpected frontend runtime errors appear during the send/receive path.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - network\n - api_diagnostic\ndiagnostics:\n - \"This is the deterministic live execution proof that sits after fixture contract and live install.\"\n - \"If the runner id mismatch is reported, rerun ensure-qa-agent-runner-pipeline.mjs --write-env.\"\nsuccess_patterns:\n - \"QA_RUNNER_OK:hello-live\"\nfailure_patterns:\n - \"plugin:qa/agent-runner/default execution failed\"\n - \"Action invoke_llm_stream call timed out\"\n - \"Agent runner temporarily unavailable\"\ntroubleshooting:\n - plugin-runtime-timeout\n","skills/skills/langbot-testing/cases/agent-runner-release-preflight.yaml":"id: agent-runner-release-preflight\ntitle: \"Agent runner release gate preflight validates environment readiness\"\nmode: agent-browser\narea: release\ntype: smoke\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - release-gate\n - preflight\n - environment\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nenv_any:\n - LANGBOT_LOCAL_AGENT_PIPELINE_URL|LANGBOT_LOCAL_AGENT_PIPELINE_NAME\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/agent-runner-release-preflight.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\nautomation_env_any:\n - LANGBOT_LOCAL_AGENT_PIPELINE_URL|LANGBOT_LOCAL_AGENT_PIPELINE_NAME\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\npreconditions:\n - \"LANGBOT_LOCAL_AGENT_PIPELINE_URL or LANGBOT_LOCAL_AGENT_PIPELINE_NAME points to the local-agent release pipeline.\"\n - \"LANGBOT_ACP_RUNNER_PIPELINE_URL or LANGBOT_ACP_RUNNER_PIPELINE_NAME points to the ACP Runner release pipeline.\"\n - \"The active browser profile is authenticated for the same LangBot backend.\"\n - \"By default the preflight performs a cheap model test for the local-agent primary model; set LANGBOT_PREFLIGHT_TEST_MODELS=0 only when deliberately classifying model credentials outside this run.\"\nsteps:\n - \"Open LANGBOT_FRONTEND_URL with the configured browser profile.\"\n - \"Use the browser token to call LangBot backend readiness APIs without printing token values.\"\n - \"Check plugin runtime status, Box status, required runner plugins, qa-plugin-smoke, and qa_plugin_echo.\"\n - \"Resolve the local-agent and ACP Runner QA pipelines from their case-specific env vars.\"\n - \"Assert each pipeline uses the expected runner id.\"\n - \"Assert the external runner pipeline uses the expected runner id.\"\n - \"Assert the local-agent primary model advertises func_call and vision for the full release gate.\"\n - \"Run the local-agent primary model test endpoint unless LANGBOT_PREFLIGHT_TEST_MODELS=0.\"\nchecks:\n - \"API diagnostic: api-diagnostic.json has no blockers and no env_issues.\"\n - \"API diagnostic: required pipelines resolve to plugin:langbot-team/LocalAgent/default and plugin:langbot-team/ACPRunner/default.\"\n - \"API diagnostic: qa_plugin_echo is exposed by /api/v1/tools.\"\n - \"API diagnostic: local-agent model check catches invalid credentials or missing func_call/vision before release E2E starts.\"\n - \"Secret safety: token values, api keys, and provider secrets are not printed.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - network\n - api_diagnostic\ndiagnostics:\n - \"blocked means the test instance is not configured for the full release gate: missing pipeline, wrong runner id, or missing plugin.\"\n - \"env_issue means the runtime or upstream dependency is not usable: backend unavailable, plugin runtime down, Box down, or the local-agent model cannot pass a model test.\"\n - \"If qa_mcp_echo is absent here, continue to mcp-stdio-register before mcp-stdio-tool-call; qa_mcp_echo is not required before registration.\"\n - \"If the model check fails with invalid api key, switch the local-agent release pipeline to a known-good func_call model before diagnosing runner code.\"\nsuccess_patterns:\n - \"Release gate preflight passed\"\nfailure_patterns:\n - \"Preflight blocked\"\n - \"Preflight environment issue\"\n - \"invalid api key\"\n - \"runner.llm_error\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - local-agent-model-route-unavailable\n - proxy-env-mismatch\n","skills/skills/langbot-testing/cases/agent-runner-runtime-chaos.yaml":"id: agent-runner-runtime-chaos\ntitle: \"Runner SDK runtime chaos pytest probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - pytest\n - runtime\n - sdk\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-runtime-chaos.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-runtime-chaos --dry-run` first; remove `--dry-run` after checking the SDK repo target and evidence directory.\"\n - \"Automation resolves LANGBOT_PLUGIN_SDK_REPO, defaulting to ../../langbot-plugin-sdk when the env var is unset.\"\n - \"Automation runs the existing SDK pytest files tests/runtime/plugin/test_mgr_runner.py and tests/runtime/test_pull_api_handlers.py.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"pytest exit status is 0 for the existing Runner runtime and pull API handler tests.\"\n - \"pytest-stdout.log and pytest-stderr.log are written under LBS_EVIDENCE_DIR.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe does not open the WebUI; it runs SDK pytest targets directly.\"\n - \"env_issue means LANGBOT_PLUGIN_SDK_REPO/default ../../langbot-plugin-sdk did not resolve, rtk/uv was unavailable, or the expected test files are missing.\"\n - \"fail means the existing SDK runtime pytest target failed or timed out.\"\nsuccess_patterns:\n - \"pytest passed\"\nfailure_patterns:\n - \"pytest exited with status\"\n - \"pytest timed out\"\n - \"Failed to start pytest command\"\n","skills/skills/langbot-testing/cases/bot-event-routing-product-flow.yaml":"","skills/skills/langbot-testing/cases/box-mcp-heartbeat-recovery.yaml":"","skills/skills/langbot-testing/cases/claude-code-agent-debug-chat.yaml":"","skills/skills/langbot-testing/cases/codex-agent-debug-chat.yaml":"","skills/skills/langbot-testing/cases/dify-agent-debug-chat.yaml":"","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-cross-pipeline-isolation.yaml":"","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-fault-recovery.yaml":"","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-load.yaml":"","skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-slow-load.yaml":"","skills/skills/langbot-testing/cases/langbot-fault-taxonomy-contract.yaml":"","skills/skills/langbot-testing/cases/langbot-live-backend-latency.yaml":"","skills/skills/langbot-testing/cases/langbot-live-backend-log-health.yaml":"","skills/skills/langbot-testing/cases/langbot-live-control-plane-api.yaml":"","skills/skills/langbot-testing/cases/langbot-overhead-accounting-contract.yaml":"","skills/skills/langbot-testing/cases/langbot-space-debug-chat-concurrency-smoke.yaml":"","skills/skills/langbot-testing/cases/langrag-kb-retrieve.yaml":"","skills/skills/langbot-testing/cases/langrag-parser-golden-e2e.yaml":"","skills/skills/langbot-testing/cases/langrag-sentinel-kb-discover.yaml":"","skills/skills/langbot-testing/cases/local-agent-basic-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-combo-rag-compaction-tool-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-complex-coding-task-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-context-compaction-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-effective-prompt-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-model-fallback-before-first-chunk-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-multimodal-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-multitool-rag-compaction-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-nonstreaming-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-parallel-tools-rag-compaction-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-plugin-tool-call-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-rag-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-rag-multimodal-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-steering-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-streaming-post-commit-failure-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-tool-error-recovery-debug-chat.yaml":"","skills/skills/langbot-testing/cases/local-agent-tool-loop-limit-debug-chat.yaml":"","skills/skills/langbot-testing/cases/mcp-stdio-register.yaml":"","skills/skills/langbot-testing/cases/mcp-stdio-tool-call.yaml":"","skills/skills/langbot-testing/cases/pipeline-debug-chat-performance.yaml":"","skills/skills/langbot-testing/cases/pipeline-debug-chat.yaml":"","skills/skills/langbot-testing/cases/plugin-e2e-smoke.yaml":"","skills/skills/langbot-testing/cases/provider-deepseek.yaml":"","skills/skills/langbot-testing/cases/qa-plugin-smoke-live-install.yaml":"","skills/skills/langbot-testing/cases/sandbox-skill-authoring-e2e.yaml":"","skills/skills/langbot-testing/cases/sandbox-skill-authoring-edit-existing-e2e.yaml":"","skills/skills/langbot-testing/cases/skill-discovery-via-mcp-gateway.yaml":"","skills/skills/langbot-testing/cases/webui-login-state.yaml":"","skills/skills/langbot-testing/cases/wizard-onebot-agent-runtime.yaml":"","skills/skills/langbot-testing/cases/wizard-runner-marketplace-catalog.yaml":"","skills/skills/langbot-testing/cases/wizard-scenario-routing.yaml":"","skills/skills/langbot-testing/cases/workspace-compatibility-preflight.yaml":"","skills/skills/langbot-testing/cases/workspace-plugin-pages-smoke.yaml":"","skills/skills/langbot-testing/cases/workspace-repository-contracts.yaml":"","skills/skills/langbot-testing/references/agent-runner-qa-workflow.md":"","skills/skills/langbot-testing/references/agent-runner-release-gate.md":"","skills/skills/langbot-testing/references/dify-agent-runner.md":"","skills/skills/langbot-testing/references/langrag-knowledge-base.md":"","skills/skills/langbot-testing/references/local-agent-runner-coverage.md":"","skills/skills/langbot-testing/references/local-agent-runner.md":"","skills/skills/langbot-testing/references/mcp-stdio-testing.md":"","skills/skills/langbot-testing/references/model-provider-testing.md":"","skills/skills/langbot-testing/references/performance-reliability-testing.md":"","skills/skills/langbot-testing/references/pipeline-debug-chat.md":"","skills/skills/langbot-testing/references/plugin-e2e-smoke.md":"","skills/skills/langbot-testing/references/sandbox-skill-authoring.md":"","skills/skills/langbot-testing/references/skill-all-tool-acceptance.md":"","skills/skills/langbot-testing/references/troubleshooting.md":"","skills/skills/langbot-testing/references/web-ui-testing.md":"","skills/skills/langbot-testing/references/workspace-release-testing.md":"","skills/skills/langbot-testing/suites/agent-runner-release-gate.yaml":"","skills/skills/langbot-testing/suites/core-smoke.yaml":"","skills/skills/langbot-testing/suites/eba-release-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-debug-chat-isolation-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-debug-chat-load-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-live-backend-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-performance-contract-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-performance-reliability-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-user-path-performance-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-workspace-contract-gate.yaml":"","skills/skills/langbot-testing/suites/langbot-workspace-release-gate.yaml":"","skills/skills/langbot-testing/suites/local-agent-gate.yaml":"","skills/skills/langbot-testing/troubleshooting/agent-runner-actor-context-fields.yaml":"","skills/skills/langbot-testing/troubleshooting/aiosqlite-connect-hangs.yaml":"","skills/skills/langbot-testing/troubleshooting/ambiguous-runner-default-label.yaml":"","skills/skills/langbot-testing/troubleshooting/backend-not-listening.yaml":"","skills/skills/langbot-testing/troubleshooting/box-runtime-silent-action-timeout.yaml":"","skills/skills/langbot-testing/troubleshooting/box-session-conflict-logical-metadata.yaml":"","skills/skills/langbot-testing/troubleshooting/debug-chat-history-contaminates-automation.yaml":"","skills/skills/langbot-testing/troubleshooting/dynamic-form-missing-config-id.yaml":"","skills/skills/langbot-testing/troubleshooting/e2b-extra-mount-sync-missing.yaml":"","skills/skills/langbot-testing/troubleshooting/local-agent-mcp-text-result-empty-follow-up.yaml":"","skills/skills/langbot-testing/troubleshooting/local-agent-model-route-unavailable.yaml":"","skills/skills/langbot-testing/troubleshooting/marketplace-network-flaky.yaml":"","skills/skills/langbot-testing/troubleshooting/mcp-stdio-args-not-applied.yaml":"","skills/skills/langbot-testing/troubleshooting/nsjail-cli-compatibility.yaml":"","skills/skills/langbot-testing/troubleshooting/pipeline-form-controlled-warning.yaml":"","skills/skills/langbot-testing/troubleshooting/plugin-dependency-install-offline.yaml":"","skills/skills/langbot-testing/troubleshooting/plugin-runtime-timeout.yaml":"","skills/skills/langbot-testing/troubleshooting/provider-image-parse-error.yaml":"","skills/skills/langbot-testing/troubleshooting/proxy-env-mismatch.yaml":"","skills/skills/langbot-testing/troubleshooting/sandbox-native-tools-unavailable.yaml":"","skills/skills/langbot-testing/troubleshooting/socks-proxy-without-socksio.yaml":"","skills/skills/langbot-testing/troubleshooting/survey-widget-blocks-debug-chat.yaml":"","skills/skills/langbot-testing/troubleshooting/telemetry-proxy-noise.yaml":"","skills/skills/langbot-testing/troubleshooting/tool-name-collision-between-mcp-and-plugin.yaml":"","skills/skills/langbot-testing/troubleshooting/uv-run-resyncs-local-sdk.yaml":""},"items":[{"name":"AGENTS.md","path":"AGENTS.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/AGENTS.md","title":"AI Agent Protocol & Instructions","category":"root-instruction","format":"markdown","content":"# AGENTS.md\n\nThis file guides code agents working in the LangBot main repository. `CLAUDE.md` is a symlink to this file.\n\nRead `ARCHITECTURE.md` before non-trivial backend, frontend, runtime, plugin, Box, MCP, persistence, or cross-repo SDK changes. This file is the working checklist; `ARCHITECTURE.md` is the system map.\n\n## Quick Facts\n\n- Python backend: `>=3.11,<4.0`, dependencies managed by `uv`.\n- Frontend: `web/` is Vite + React Router 7 + shadcn/ui + Tailwind, managed by `pnpm`.\n- Backend framework: Quart served by Hypercorn on `api.port`, default `5300`.\n- Frontend dev server: `web/` on `3000`, with `VITE_API_BASE_URL` pointing at the backend.\n- Plugin/Box/runtime contracts live in sibling repo `langbot-plugin-sdk`, pinned as `langbot-plugin` in `pyproject.toml`.\n\n## Essential Commands\n\n```bash\nuv sync --dev\nuv run main.py\nuv run pre-commit install\n\ncd web\npnpm install\npnpm dev\npnpm build\n```\n\nUseful focused tests:\n\n```bash\nuv run pytest tests/unit_tests -q\nuv run pytest tests/integration -q\nuv run pytest tests/integration/persistence -q\nuv run pytest tests/manual/mcp_smoke.py\n\ncd web\npnpm lint\npnpm test:e2e\n```\n\nRun the narrowest useful test first, then broader checks when confidence is needed.\n\n## Where to Look\n\n- Architecture map: `ARCHITECTURE.md`.\n- Dev environment guide: https://langbot.app/docs/zh/develop/dev-config.\n- Plugin runtime / CLI / SDK debugging: https://langbot.app/docs/zh/develop/plugin-runtime.\n- API-key auth: `docs/API_KEY_AUTH.md`.\n- Service API CLI: [`langbot-cli`](https://github.com/langbot-app/langbot-cli) (`lbctl`) manages running Workspaces; it is separate from the SDK's `lbp` CLI.\n- Box deep-dive notes: `docs/review/box-architecture.md` and related files.\n- In-repo skills: `skills/` is the single source of truth for LangBot agent skills.\n- SDK repo: `../langbot-plugin-sdk/` when changing shared entities, plugin APIs, action protocol, `lbp rt`, or `lbp box`.\n\n## Cross-Repo SDK Work\n\nWhen changing SDK contracts used by LangBot:\n\n```bash\n# from langbot-plugin-sdk, with LangBot's .venv active\nuv pip install .\n\n# from LangBot, preserve the locally installed SDK\nuv run --no-sync main.py\n```\n\nFor standalone runtime debugging:\n\n```bash\n# in langbot-plugin-sdk\nuv run --no-sync lbp rt\nuv run --no-sync lbp box\n\n# in LangBot\nuv run --no-sync main.py --standalone-runtime\nuv run --no-sync main.py --standalone-box\n```\n\nConfig keys to verify in `data/config.yaml` / `src/langbot/templates/config.yaml`:\n\n- Plugin runtime: `plugin.runtime_ws_url`, default Docker host `langbot_plugin_runtime:5400/control/ws`.\n- Box runtime: `box.enabled`, `box.backend`, `box.runtime.endpoint`, Docker host `langbot_box:5410`.\n- API/MCP auth: `api.global_api_key`.\n\n## Change Rules\n\n- HTTP API changes that should be agent-accessible must update the matching MCP tool in `src/langbot/pkg/api/mcp/server.py` and the relevant skill under `skills/` in the same pass.\n- When changing Service API routes or capabilities used by `lbctl`, check compatibility with the separate `langbot-cli` repository.\n- New schema changes use Alembic under `src/langbot/pkg/persistence/alembic/versions/`. LangBot 4.x does not support upgrading 3.x databases.\n- New platform behavior belongs in platform adapters only for platform translation; pipeline/business logic belongs in `pkg/pipeline/` or services.\n- User-facing strings must support i18n; follow the backend language rules below for server-generated content.\n- Conversation payload columns, including new ones, must use `MediaReferenceText` and persist image references only. See the [media persistence contract](ARCHITECTURE.md#persistence-and-configuration).\n- Interactive session lists use `SessionSummaryReader` to read only `monitoring_sessions`; `SessionSummaryWriter` maintains summaries in the source transaction. Historical initialization completes before writers start, never in requests or application background tasks. Follow the [conversation summary contract](ARCHITECTURE.md#conversation-summary-boundary) and keep `uv run pytest --required-suite session_summary` passing.\n- Code comments and docstrings must be English.\n- Keep compatibility and security in mind; LangBot is used in both self-hosted/community and toB deployments.\n- Commit message format: `(): `.\n\n## Backend Message Language\n\n- LangBot-owned business errors, runtime notices, and default messages sent to users must use the owning Workspace's message language. Keep this independent of each member's WebUI language; an unset Workspace language defaults to `en_US`.\n- Reuse `src/langbot/pkg/i18n/` and the existing `message_language` entry in `WorkspaceMetadata`. Use `t()` with stable English templates and named parameters; add matching keys and placeholders to all eight catalogs: `en_US`, `zh_Hans`, `zh_Hant`, `ja_JP`, `es_ES`, `ru_RU`, `th_TH`, and `vi_VN`. Avoid new translation frameworks or duplicate language settings.\n- Resolve language from a trusted Workspace or bot execution context, and isolate it across concurrent requests, tasks, and adapter callbacks. Browser language hints may supply account-only or unauthenticated response language, but must not override a resolved Workspace or authorize tenant selection. Reading/changing the setting requires `workspace.view` / `workspace.update`.\n- Pipeline notices should be configurable in Pipeline output settings through `pipeline_notice()`. Blank overrides inherit the Workspace translation. Preserve user-authored text literally except for documented placeholders, and retain existing visibility switches. Do not rewrite historical messages or custom text when the language changes.\n- China-only adapters and configuration/field validation diagnostics do not require localization. An adapter also serving international users is not exempt merely because it supports a Chinese platform.\n- Preserve raw provider/plugin diagnostics, tracebacks, protocol values, and machine-readable error/status codes. Translate LangBot's own category or explanation around them; do not translate arbitrary errors with substring replacement. Plugin-authored messages remain the plugin's localization responsibility.\n- Keep changes small and maintainable. Reuse existing persistence and response boundaries; delete obsolete fields only after verifying they have no remaining readers, including compatibility paths.\n- When changing language behavior, verify catalog/placeholder parity, Workspace isolation and permissions, and Pipeline inheritance/custom overrides. Update affected configuration snapshots and API/MCP documentation. Implementation details and focused test locations: [Backend message language](docs/architecture/backend-i18n.md).\n\n## Runtime Pitfalls\n\n- Local stdio Plugin Runtime disconnects do not auto-reconnect; restart LangBot if that path breaks.\n- Orphan runtime processes on `5400`/`5401` commonly break plugin debugging.\n- Use `uv run --no-sync` after locally installing the SDK, or `uv` may restore the pinned package.\n- A false Box “no backend” often means Docker is running but the current user lacks Docker socket permission.\n- Do not confuse external MCP servers LangBot connects to (`pkg/provider/tools/loaders/mcp.py`) with LangBot's own `/mcp` server (`pkg/api/mcp/`).\n- `CLAUDE.md` is a symlink to this file; edit `AGENTS.md`, not the symlink.\n\n## Principles\n\n- Keep it simple, stupid.\n- Entities should not be multiplied unnecessarily.\n- 八荣八耻\n\n 以瞎猜接口为耻，以认真查询为荣。\n 以模糊执行为耻，以寻求确认为荣。\n 以臆想业务为耻，以人类确认为荣。\n 以创造接口为耻，以复用现有为荣。\n 以跳过验证为耻，以主动测试为荣。\n 以破坏架构为耻，以遵循规范为荣。\n 以假装理解为耻，以诚实无知为荣。\n 以盲目修改为耻，以谨慎重构为荣。\n","isInternal":false,"tokens":1847,"sizeBytes":7848},{"name":"AGENTS.md","path":"skills/AGENTS.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/AGENTS.md","title":"skills - AI Agent Protocol & Instructions","category":"root-instruction","format":"markdown","content":"# Agent Workflow\n\nThis repository stores reusable LangBot agent-testing assets. Keep changes structured so the next agent does not need to rediscover paths.\n\n## First Steps\n\n1. Read `skills/.env` before using local URLs, paths, browser profiles, or proxy defaults. If present, `skills/.env.local` overrides it for this machine and must not be committed. On a new machine, copy `skills/.env.example` to `skills/.env.local` first.\n2. Pick the smallest relevant skill:\n - `langbot-env-setup` for environment, browser, OAuth, proxy, and startup.\n - `langbot-testing` for WebUI, provider, pipeline, cases, and troubleshooting.\n - `langbot-skills-maintenance` for adding, deduplicating, or auditing this skills repository.\n3. Prefer existing cases and troubleshooting entries before exploring from scratch.\n\n## Editing Rules\n\n- UI/browser testing is the primary QA path. API/curl checks are diagnostic only and cannot make a UI case pass by themselves.\n- Put skills under `skills//`.\n- Keep `SKILL.md` concise; move detailed workflows to `references/`.\n- Put reusable test paths in `cases/*.yaml`.\n- New or edited cases must include `priority`, `risk`, `ci_eligible`, and `evidence_required` so agents can select the right test set without rereading every file.\n- Use `env_any` / `automation_env_any` for one-of machine inputs, such as `LANGBOT_PIPELINE_URL|LANGBOT_PIPELINE_NAME`; do not list those alternatives as separate all-required env keys.\n- Put reusable groups of cases in `suites/*.yaml` rather than hardcoding test sets in docs or CLI code.\n- Put growing failure knowledge in `troubleshooting/*.yaml`.\n- Do not hardcode local ports in testing docs; use `skills/.env` variables and machine-local `skills/.env.local` overrides.\n- Do not store secrets, API keys, OAuth tokens, or localStorage token values.\n\n## Required Checks\n\nAfter structural changes, run:\n\n```bash\nbin/lbs validate\n```\n\nAfter changing skills, cases, or troubleshooting assets, run:\n\n```bash\nbin/lbs index\n```\n\nUse `bin/lbs env show` to inspect defaults and `bin/lbs env doctor` when diagnosing local environment readiness. Env output is redacted by default; do not work around that by printing raw secrets.\n`bin/lbs` is a generated local wrapper. If it is missing on a fresh checkout, run `npm run bootstrap` from this directory first; `npm install` also regenerates it via `prepare`.\nUse `bin/lbs fixture check` before fixture-heavy cases such as MCP, RAG, multimodal, or plugin smoke tests.\nUse `bin/lbs case list --ready` for cases that have no missing machine inputs and no manual preconditions. Use `bin/lbs case list --machine-ready` when you want to keep `manual-check` candidates and confirm their preconditions yourself.\n\nBefore executing a saved QA path, generate the agent-facing plan:\n\n```bash\nbin/lbs test plan \n```\n\nRead the plan readiness sections before running the browser path. Missing env,\nautomation env, or fixture readiness means the case is not ready to execute and\nshould be marked `blocked` or fixed first.\n`manual_check` means machine inputs are present but the agent must verify the\ndeclared `preconditions` or `setup` items before executing the UI path. Do not\nturn a `manual_check` case into `pass` until those items were checked in the\nsame run.\n\nBefore executing a group of saved QA paths, generate the suite plan:\n\n```bash\nbin/lbs suite plan \n```\n\nUse `bin/lbs suite start ` to create a shared suite run id, suite evidence root, per-case evidence directories, and `suite-start.json`/`suite-start.md` handoff files. Then run `bin/lbs suite report --evidence-dir ` to aggregate case results.\nAutomation scripts write `automation-result.json`; write the final per-case `result.json` with `bin/lbs test result --result --reason --evidence-dir --evidence ` after collecting the required evidence. A `pass` result must include all required evidence.\nFor runner-specific Debug Chat cases, prefer case-specific pipeline env keys such as `LANGBOT_LOCAL_AGENT_PIPELINE_URL` over the generic `LANGBOT_PIPELINE_URL`; otherwise an agent can accidentally test the wrong runner.\n","isInternal":false,"tokens":957,"sizeBytes":4169},{"name":"CLAUDE.md","path":"CLAUDE.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/CLAUDE.md","title":"Claude Agent Guidelines & System Prompt","category":"claude-rule","format":"markdown","content":"AGENTS.md","isInternal":false,"tokens":3,"sizeBytes":9},{"name":"user-guide.md","path":"skills/docs/user-guide.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/docs/user-guide.md","title":"Docs Skill","category":"anthropic-skill","format":"markdown","content":"# LangBot QA Skills User Guide\n\nUse this guide as the first operational path after reading `README.md` and\n`AGENTS.md`.\n\n## 1. Configure Local Inputs\n\nRead `skills/.env`, then create `skills/.env.local` for machine-local values.\nDo not commit `.env.local`, browser profiles, reports, tokens, API keys, OAuth\nstate, or provider credentials.\n\nMinimum local fields for live browser QA:\n\n```bash\nLANGBOT_REPO=/path/to/LangBot\nLANGBOT_WEB_REPO=/path/to/LangBot/web\nLANGBOT_BACKEND_URL=http://127.0.0.1:5300\nLANGBOT_FRONTEND_URL=http://127.0.0.1:3000\nLANGBOT_DEV_FRONTEND_URL=http://127.0.0.1:3000\nLANGBOT_BROWSER_PROFILE=/path/to/langbot-browser-profile\nLANGBOT_CHROMIUM_EXECUTABLE=/path/to/chromium-or-playwright-chrome\nLANGBOT_E2E_LOGIN_USER=qa-local@example.com\n```\n\n`LANGBOT_E2E_LOGIN_USER` is a local QA account. The setup automation uses the\nLangBot recovery key from the active checkout to initialize or refresh that\nlocal account and write a browser `localStorage` token. It does not need the\nuser's GitHub or Space credentials.\n\n## 2. Check Readiness\n\nFrom `skills/`:\n\n```bash\nbin/lbs env show\nbin/lbs env doctor\nbin/lbs validate\nbin/lbs index --check\n```\n\n`env doctor` should report reachable backend and frontend URLs before live\nbrowser cases are run. Missing Space provider credentials are not a LangBot\nproduct pass; classify them as `env_issue` and configure the local Space\nprovider before measuring Debug Chat performance.\n\n## 3. Start Services\n\nStart the backend from `LANGBOT_REPO`:\n\n```bash\ncd \"$LANGBOT_REPO\"\nuv run main.py\n```\n\nStart the standalone frontend from `LANGBOT_WEB_REPO` and point it at the\nbackend:\n\n```bash\ncd \"$LANGBOT_WEB_REPO\"\nVITE_API_BASE_URL=\"$LANGBOT_BACKEND_URL\" pnpm dev --host 0.0.0.0\n```\n\nIf `VITE_API_BASE_URL` is missing, browser tests can load the Vite page but send\nAPI requests to the frontend port, which produces false UI failures.\n\n## 4. Prepare User-Path Fixtures\n\nFor local-agent Debug Chat cases and the user-path performance gate:\n\n```bash\nnode scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\n```\n\nThe script:\n\n- refreshes the local QA login and browser token;\n- marks the local wizard as skipped;\n- creates or updates a local QA pipeline;\n- scans Space LLM models, tests candidates, and switches to the first working\n Space model with tested fallback models;\n- writes `LANGBOT_PIPELINE_URL`, `LANGBOT_PIPELINE_NAME`, and local-agent\n pipeline/model variables into `skills/.env.local`;\n- returns `env_issue` when no Space model can be scanned or tested.\n\nUseful model controls:\n\n```bash\nLANGBOT_E2E_MODEL_TEST_LIMIT=8\nLANGBOT_E2E_MODEL_FALLBACK_COUNT=3\nLANGBOT_E2E_SKIP_MODEL_UUIDS=uuid-a,uuid-b\nLANGBOT_E2E_SKIP_MODEL_NAMES=model-a,model-b\nLANGBOT_E2E_SCAN_SPACE_MODELS=true\n```\n\nThe setup writes a current-runtime compatibility `max-round` value into the\npipeline config because this backend still reads that field directly during\nmessage truncation. Do not treat it as a long-term QA contract.\n\n## 5. Run Gates\n\nCross-repository contract gate (no model calls):\n\n```bash\nbin/lbs suite plan langbot-workspace-contract-gate\nbin/lbs suite run langbot-workspace-contract-gate --run-id langbot-workspace-contract-local\n```\n\nTop-down workspace release gate (browser workflows plus one complex Agent task):\n\n```bash\nbin/lbs suite plan langbot-workspace-release-gate\nbin/lbs suite run langbot-workspace-release-gate --run-id langbot-workspace-release-local --include-manual-check\n```\n\nRun `agent-run-ledger-audit` immediately after the complex Agent case. Set\n`LANGBOT_AGENT_RUN_ID` when the test instance has concurrent operator traffic.\n\nFast contract gate, no live service required:\n\n```bash\nbin/lbs suite run langbot-performance-contract-gate --run-id langbot-contract-local\n```\n\nLive backend gate:\n\n```bash\nbin/lbs suite run langbot-live-backend-gate --run-id langbot-backend-local\n```\n\nBrowser-visible user-path performance gate:\n\n```bash\nbin/lbs suite plan langbot-user-path-performance-gate\nbin/lbs suite run langbot-user-path-performance-gate --run-id langbot-user-path-local --include-manual-check\n```\n\nControlled Debug Chat message-path load gate (manual/non-required; run fake-provider cases serially when they share `LANGBOT_FAKE_PROVIDER_URL`):\n\n```bash\nbin/lbs suite plan langbot-debug-chat-load-gate\nbin/lbs test run langbot-fake-provider-debug-chat-load --run-id langbot-fake-load-local\nbin/lbs test run langbot-fake-provider-debug-chat-slow-load --run-id langbot-fake-slow-local\nbin/lbs test run langbot-fake-provider-debug-chat-fault-recovery --run-id langbot-fake-fault-local\nbin/lbs test run langbot-space-debug-chat-concurrency-smoke --run-id langbot-space-smoke-local\n```\n\nCross-pipeline Debug Chat isolation is a separate manual regression gate because\ncurrent releases may fail it due to product bug #2286:\n\n```bash\nbin/lbs suite plan langbot-debug-chat-isolation-gate\nbin/lbs suite run langbot-debug-chat-isolation-gate --run-id langbot-debug-chat-isolation-local --include-manual-check\n```\n\nStart with `langbot-fake-provider-debug-chat-load`. It launches a local\nOpenAI-compatible fake provider, creates the matching provider/model/pipeline,\nthen sends concurrent WebSocket Debug Chat messages through the real backend.\nUse `langbot-fake-provider-debug-chat-slow-load` to measure the same path under\ndeterministic streaming latency. Use\n`langbot-fake-provider-debug-chat-fault-recovery` to inject bounded provider\nHTTP failures and confirm later Debug Chat requests recover. Use the separate\n`langbot-debug-chat-isolation-gate` to verify that concurrent Debug Chat traffic\non two pipelines does not leak assistant responses across pipeline boundaries;\ncurrent releases may fail that gate because of #2286, so keep it out of the\nnormal load gate until the product fix lands.\nUse `langbot-space-debug-chat-concurrency-smoke` only as a low-volume live\nprovider smoke; it includes Space/model/network latency and should be compared\nagainst the fake-provider baseline before attributing failures to LangBot.\n\n`manual_check` means the agent must confirm the declared preconditions for that\nrun window. When setup automation is declared, run output may stop early with\n`env_issue`; fix that environment input before treating the product path as\nmeasured.\n\n## 6. Read Results\n\nSuite reports live under `skills/reports/`. Evidence lives under\n`skills/reports/evidence//`.\n\nFor performance cases, inspect:\n\n- `metrics.json` for p50/p95/p99, error rate, and total duration;\n- `automation-result.json` for threshold decisions and artifacts;\n- `console.log` and `network.log` for frontend/API failures;\n- backend logs for provider, runner, WebSocket, or persistence failures.\n\nDo not call a user-path performance result a LangBot overhead regression until\nprovider/tool/network time has been separated or ruled out.\n","isInternal":false,"tokens":1628,"sizeBytes":6795},{"name":"package.json","path":"skills/package.json","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/package.json","title":"Skills Skill","category":"anthropic-skill","format":"json","content":"{\n \"private\": true,\n \"type\": \"module\",\n \"bin\": {\n \"lbs\": \"./bin/lbs\"\n },\n \"scripts\": {\n \"bootstrap\": \"node scripts/bootstrap-lbs.mjs\",\n \"prepare\": \"node scripts/bootstrap-lbs.mjs\",\n \"prevalidate\": \"node scripts/bootstrap-lbs.mjs\",\n \"preindex\": \"node scripts/bootstrap-lbs.mjs\",\n \"preindex:check\": \"node scripts/bootstrap-lbs.mjs\",\n \"pretest\": \"node scripts/bootstrap-lbs.mjs\",\n \"precheck\": \"node scripts/bootstrap-lbs.mjs\",\n \"lbs\": \"node src/lbs.ts\",\n \"test\": \"node test/lbs-cli.test.ts && python3 -m unittest discover -s test -p 'test_*.py'\",\n \"validate\": \"node src/lbs.ts validate\",\n \"index\": \"node src/lbs.ts index\",\n \"index:check\": \"node src/lbs.ts index --check\",\n \"check:syntax\": \"find src test scripts -type f \\\\( -name '*.ts' -o -name '*.mjs' \\\\) -print0 | xargs -0 -n1 node --check\",\n \"check\": \"npm run check:syntax && npm run validate && npm test\"\n },\n \"engines\": {\n \"node\": \">=22.6\"\n },\n \"devDependencies\": {\n \"playwright\": \"^1.60.0\"\n }\n}\n","frontmatter":{"private":true,"type":"module","bin":{"lbs":"./bin/lbs"},"scripts":{"bootstrap":"node scripts/bootstrap-lbs.mjs","prepare":"node scripts/bootstrap-lbs.mjs","prevalidate":"node scripts/bootstrap-lbs.mjs","preindex":"node scripts/bootstrap-lbs.mjs","preindex:check":"node scripts/bootstrap-lbs.mjs","pretest":"node scripts/bootstrap-lbs.mjs","precheck":"node scripts/bootstrap-lbs.mjs","lbs":"node src/lbs.ts","test":"node test/lbs-cli.test.ts && python3 -m unittest discover -s test -p 'test_*.py'","validate":"node src/lbs.ts validate","index":"node src/lbs.ts index","index:check":"node src/lbs.ts index --check","check:syntax":"find src test scripts -type f \\( -name '*.ts' -o -name '*.mjs' \\) -print0 | xargs -0 -n1 node --check","check":"npm run check:syntax && npm run validate && npm test"},"engines":{"node":">=22.6"},"devDependencies":{"playwright":"^1.60.0"}},"isInternal":false,"tokens":319,"sizeBytes":1011},{"name":"00-technology-options.md","path":"skills/qa-agent-docs/qa-agent/00-technology-options.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/qa-agent-docs/qa-agent/00-technology-options.md","title":"Qa-agent Skill","category":"anthropic-skill","format":"markdown","content":"# LangBot Agent Testing 技术选型\n\n## 状态\n\n这是技术选型背景文档，不是当前路线图。当前黑盒 E2E QA 的实施顺序见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n## 目标\n\n`langbot-skills` 的目标不是替代测试框架，而是沉淀 agent 可复用的测试资产，让开发者 clone 仓库后，可以让 Codex、Claude Code、Computer Use 或 Playwright MCP 复用已有路径完成 LangBot 功能验证。\n\n核心原则：\n\n- Skill 负责路由和少量规则。\n- Reference 负责可读流程和背景知识。\n- Case 负责结构化测试路径。\n- Troubleshooting 负责结构化故障资产。\n- `lbs` 负责结构校验、索引、资产创建和未来的运行/报告能力。\n- UI/browser 是产品 QA 的主路径；API/curl 只用于诊断。\n\n## 浏览器控制层\n\n不同开发者可用的浏览器控制能力不同，所以浏览器层必须可替换。\n\n| 方案 | 适用场景 | 优点 | 代价 |\n|---|---|---|---|\n| Codex / Claude Computer Use | agent 可以直接控制可见浏览器 | 登录和交互路径最自然，通常不需要额外 MCP 浏览器桥接 | 依赖具体 agent 工具能力 |\n| Playwright MCP | 没有 Computer Use，但有 MCP 浏览器工具 | 稳定、可脚本化、适合回归路径 | OAuth 登录通常需要额外 visible profile |\n| 直接 Playwright 脚本 | 测试路径非常稳定，适合 CI | 可重复性强 | 需要维护脚本和 selector |\n| 商业 AI QA 平台 | 团队希望外包测试运行平台 | 报告和 PR 集成完整 | 成本和平台绑定 |\n\n## 当前推荐\n\n先采用分层降级：\n\n```text\n有 Computer Use？\n 是 -> 使用 Computer Use 控制浏览器\n 否 -> 使用 Playwright MCP\n\n需要 GitHub OAuth？\n 是 -> 使用持久浏览器 profile，让用户手动完成登录\n 否 -> 直接使用已有登录态或测试账号状态\n```\n\n具体选择逻辑沉淀在：\n\n```text\nskills/langbot-env-setup/references/browser-access-selection.md\n```\n\n测试原则固定在：\n\n```text\ndocs/qa-agent/03-agent-browser-qa-principles.md\n```\n\n## 环境变量层\n\n测试文档不应写死端口。共享默认值放在：\n\n```text\nskills/.env\n```\n\n关键变量：\n\n```text\nLANGBOT_FRONTEND_URL\nLANGBOT_BACKEND_URL\nLANGBOT_DEV_FRONTEND_URL\nLANGBOT_REPO\nLANGBOT_WEB_REPO\nLANGBOT_BROWSER_PROFILE\n```\n\nAgent 执行测试前应先读取 `skills/.env`，再用用户提供的当前环境或已启动服务覆盖默认值。\n\n## 测试资产层\n\n测试资产分两类：\n\n```text\nskills//\n references/ # Markdown 流程说明\n cases/ # 结构化测试用例\n troubleshooting/ # 结构化故障记录\n```\n\n当前已实现：\n\n- `SKILL.md` 路由\n- `references/*.md`\n- `lbs case new/list/show`\n- `lbs trouble show/search`\n- `lbs test plan`\n- `lbs test report`\n- `lbs list / validate / index`\n\n下一步重点：\n\n- 日志守卫规则补充\n- 报告产物管理\n\n## 关键判断\n\n不要强制所有内容只能通过 CLI 修改。更好的模式是：\n\n- 新增 case/troubleshooting：优先使用 `lbs`\n- 大段流程说明：允许直接编辑 Markdown\n- 结构性变更后：必须运行 `lbs validate`\n- 任何生成索引的变更后：运行 `lbs index`\n\n这样既能沉淀结构化资产，又不会在 schema 未稳定时拖慢迭代。\n","isInternal":false,"tokens":1112,"sizeBytes":3288},{"name":"01-qa-agent-harness-plan.md","path":"skills/qa-agent-docs/qa-agent/01-qa-agent-harness-plan.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/qa-agent-docs/qa-agent/01-qa-agent-harness-plan.md","title":"Qa-agent Skill","category":"anthropic-skill","format":"markdown","content":"# LangBot Skills 测试资产库规划\n\n## 状态\n\n这是早期测试资产库规划文档，保留用于解释 `langbot-skills` 的分层来源。\n\n当前路线已经收敛为黑盒 E2E QA：开发者用 agent 通过浏览器测试 LangBot，\n稳定路径沉淀为 case，失败知识沉淀为 troubleshooting。`lbs test report` 和\n日志守卫已有 MVP，后续重点是报告证据、case 元数据和少量稳定路径自动化。当前优先级见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n本文中关于 `case list/show`、`trouble show/search`、`test plan` 的“计划实现”\n内容已经部分过时，因为这些能力已经落地。\n\n## 目标\n\n让开发者 clone `langbot-skills` 后，可以把测试意图交给 agent，由 agent 复用已有环境配置、测试路径和故障知识完成 LangBot 功能验证。\n\n典型场景：\n\n- 冒烟测试：验证 pipeline Debug Chat、provider、常见页面是否正常。\n- Provider 测试：添加 DeepSeek/OpenAI/Claude 等供应商并验证模型可用。\n- 新 feature 测试：探索新 UI 路径，并在稳定后沉淀成 case/reference。\n- 回归测试：复用旧路径，避免每个窗口重新探索登录、模型配置、pipeline 调试。\n- 故障沉淀：把 runtime 超时、代理不一致、WebSocket 问题记录为可搜索资产。\n\n核心方向见 `03-agent-browser-qa-principles.md`：agent 必须以浏览器/UI 为主路径，API/curl 只能作为诊断手段。\n\n## 当前仓库结构\n\n```text\nskills/\n .env # 共享默认变量\n langbot-env-setup/ # 环境准备、浏览器控制路径、代理、登录态\n langbot-testing/ # WebUI / provider / pipeline 测试入口\n langbot-plugin-dev/ # 插件开发测试\n langbot-eba-adapter-dev/ # 平台适配器开发测试\nsrc/\n lbs.ts # CLI 源码\nbin/\n lbs # CLI 入口\ndocs/\n qa-agent/ # 规划文档，历史目录名保留\n```\n\n## 设计分层\n\n### 1. Skill 层\n\n`SKILL.md` 只做触发和路由，不承载大段流程。\n\n例子：\n\n```text\nlangbot-env-setup -> 选择 Computer Use / Playwright MCP / OAuth profile / proxy\nlangbot-testing -> 选择 WebUI / pipeline / provider / troubleshooting\n```\n\n### 2. Reference 层\n\nMarkdown 记录人和 agent 都能读的流程说明。\n\n适合内容：\n\n- 如何选择浏览器控制方式\n- 如何启动/检查服务\n- 如何执行 pipeline Debug Chat\n- 如何处理 OAuth 登录态\n\n### 3. Case 层\n\n使用 YAML 记录可重复测试路径。\n\n建议结构：\n\n```text\nskills/langbot-testing/cases/\n pipeline-debug-chat.yaml\n provider-deepseek.yaml\n```\n\n建议格式：\n\n```yaml\nid: pipeline-debug-chat\ntitle: Pipeline Debug Chat returns a bot response\nmode: agent-browser\narea: pipeline\ntype: smoke\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nsteps:\n - Open LANGBOT_FRONTEND_URL\n - Navigate to Pipelines\n - Open target pipeline\n - Select Debug Chat\n - Send deterministic prompt\nchecks:\n - \"UI: User message appears\"\n - \"UI: Bot message appears\"\n - \"Console: No unexpected frontend errors\"\n - \"Logs: Backend log includes Conversation(0) Streaming completed\"\ndiagnostics:\n - \"Use API/curl only after the UI path is attempted, to distinguish frontend display failure from backend/runtime failure.\"\ntroubleshooting:\n - plugin-runtime-timeout\n - proxy-env-mismatch\n```\n\n### 4. Troubleshooting 层\n\n故障资产会逐渐变大，适合结构化记录。\n\n历史 Markdown 入口保留在：\n\n```text\nskills/langbot-testing/references/troubleshooting.md\n```\n\n当前 canonical 结构化故障资产在：\n\n```text\nskills/langbot-testing/troubleshooting/\n plugin-runtime-timeout.yaml\n proxy-env-mismatch.yaml\n```\n\n### 5. CLI 层\n\n`lbs` 是统一入口，不再引入独立 `qa` 命令。\n\n已实现或当前可用：\n\n```bash\nbin/lbs list\nbin/lbs validate\nbin/lbs index\nbin/lbs new-skill \nbin/lbs new-ref \nbin/lbs case new pipeline-debug-chat --title \"Pipeline Debug Chat\"\nbin/lbs case list\nbin/lbs case show pipeline-debug-chat\nbin/lbs trouble list \nbin/lbs trouble show plugin-runtime-timeout\nbin/lbs trouble search runtime\nbin/lbs trouble add --title ... --symptom ... --cause ... --fix ...\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log\n```\n\n## 测试库位置\n\n不要使用隐藏 `.qa/` 作为主测试库。测试资产应该和 skill 放在一起，便于触发和维护：\n\n```text\nskills/langbot-testing/\n references/\n cases/\n troubleshooting/\n reports/ # 可选，本地运行产物可按需忽略或输出到外部目录\n```\n\n如果未来需要项目本地测试库，可以允许 `lbs` 支持 `--workspace` 或项目根目录配置，但 canonical 资产仍保存在 `langbot-skills`。\n\n## 阶段规划\n\n### 阶段一：环境和测试路径沉淀\n\n状态：基本完成，持续维护。\n\n- `skills/.env` 管共享默认变量。\n- `langbot-env-setup` 拆出 Computer Use、Playwright MCP、OAuth profile、proxy、service startup。\n- `langbot-testing` 记录 WebUI、pipeline、provider 测试路径。\n- `lbs validate/index` 维护结构。\n\n完成标准：\n\n- agent 可以从 `skills/.env` 和 references 中找到当前测试入口。\n- pipeline Debug Chat 这类路径不再需要从头探索。\n\n### 阶段二：结构化 case/troubleshooting\n\n状态：主体已完成，继续补齐元数据和资产质量。\n\n目标：\n\n- `lbs case new/list/show`\n- `lbs trouble show/search`\n- case id 去重、字段校验、索引生成\n\n完成标准：\n\n- 冒烟测试路径可以用结构化 case 表示。\n- 下一个 agent 窗口可以直接读取 case 执行。\n\n### 阶段三：计划和报告\n\n状态：已有 MVP，继续完善。\n\n目标：\n\n- `lbs test plan `\n- agent 按 plan 使用浏览器执行 UI QA\n- `lbs test report`\n- 日志守卫集成\n- 报告产物和 evidence 约定\n\n完成标准：\n\n- agent 可以按 case plan 执行浏览器测试。\n- 结果报告包含 UI 结果、后端日志、console 错误和 troubleshooting 建议。\n\n## 执行规则\n\n- agent 可以直接编辑 Markdown reference。\n- 新增结构化 case/troubleshooting 时，优先使用 `lbs`。\n- 每次结构变更后运行 `bin/lbs validate`。\n- 每次索引相关变更后运行 `bin/lbs index`。\n- 测试文档不写死端口，使用 `skills/.env` 中的 URL 变量。\n- 测试 case 的 `mode` 固定为 `agent-browser`。\n- API/curl 只能写入 `diagnostics`，不能替代 UI 步骤和 UI 检查。\n","isInternal":false,"tokens":2067,"sizeBytes":6765},{"name":"02-log-guard-plan.md","path":"skills/qa-agent-docs/qa-agent/02-log-guard-plan.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/qa-agent-docs/qa-agent/02-log-guard-plan.md","title":"Qa-agent Skill","category":"anthropic-skill","format":"markdown","content":"# 日志守卫规划\n\n## 状态\n\n这是当前活跃设计，已有第一版文件扫描 MVP。实现边界需要和黑盒 E2E 路线保持一致：\n\n- 日志守卫服务于 `lbs test report`。\n- 它不替代浏览器/UI 判断。\n- 它不发展成独立后端 API 测试框架。\n- 第一版默认扫描 `LANGBOT_REPO/data/logs/` 下最新的 `langbot-*.log`，也可扫描 agent\n 显式提供的 backend/frontend/console 日志文件。\n\n当前总体路线见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n## 目标\n\n日志守卫是 `lbs test report` 的一部分，用来在 agent 执行测试期间捕获 UI 断言之外的运行时问题。\n\n当前命令方向已收敛为 `lbs test plan` / `lbs test report`。日志守卫服务于 agent-browser QA，不是独立的后端 API 测试入口。\n\nLangBot 是异步且集成度高的系统，有些问题不会直接表现为页面失败：\n\n- 后台任务异常\n- 未等待的协程\n- Provider 流式调用失败\n- 插件 runtime 超时\n- 平台发送失败\n- 数据库连接问题\n- 敏感信息泄露\n\n日志守卫负责把这些信号结构化地放进测试报告，并关联到 troubleshooting 资产。\n\n## 输入\n\n日志守卫应从环境和运行上下文读取配置：\n\n- `skills/.env` 中的 `LANGBOT_BACKEND_URL`\n- `skills/.env` 中的 `LANGBOT_REPO`，用于自动发现 LangBot 后端日志\n- `lbs test plan` / report 记录的 case id\n- LangBot 后端进程输出\n- 前端 dev server 输出\n- 浏览器 console/network 错误\n- case 声明的 success/failure patterns 和 expected failures\n\n## MVP 范围\n\n- 读取一个或多个日志流或日志文件。\n- 检测错误模式。\n- 支持按 case id 或 pattern 白名单。\n- 输出 JSON/Markdown 摘要。\n- 发现非预期错误时让测试报告标记失败；未来如果有自动执行器，再返回非零退出码。\n\n## 错误分类\n\n### 永远非预期\n\n除非 case 明确声明，否则应失败：\n\n- `Traceback`\n- `Task exception was never retrieved`\n- `RuntimeWarning: coroutine .* was never awaited`\n- `Unclosed client session`\n- `Unclosed connector`\n- `KeyError`\n- `TypeError`\n- `AttributeError`\n- 密钥、token、secret 明文泄露\n\n### Case 预期错误\n\n只有当前 case 声明时允许：\n\n- 无效 provider key\n- Provider 认证失败\n- 无效 webhook payload\n- 插件测试故意抛错\n- 超时测试\n- 限流测试\n\n### 仅警告\n\n报告但默认不失败：\n\n- 可恢复重试\n- 恢复的超时\n- 废弃配置\n- 慢请求\n- 版本检查失败\n\n## 与 Troubleshooting 集成\n\n日志守卫不只输出错误文本，还应尽量匹配已知 troubleshooting id。\n\n例子：\n\n```text\nAction list_plugins call timed out\nAction list_runners call timed out\nAction invoke_llm_stream call timed out\n```\n\n可映射到：\n\n```text\nplugin-runtime-timeout\n```\n\n```text\nuppercase proxy points to one host, lowercase proxy points to another\n```\n\n可映射到：\n\n```text\nproxy-env-mismatch\n```\n\n## 未来命令\n\n```bash\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report --output report.md\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log --console-log /path/to/console.log\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs test report pipeline-debug-chat --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\" --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --no-auto-log\n```\n\n运行报告应包含：\n\n- case id\n- URL 和环境变量摘要，不能包含 secrets\n- 浏览器可见结果\n- 后端日志摘要\n- console/network 错误\n- 匹配到的 troubleshooting id\n- 通过/失败结论\n\n## MVP 完成标准\n\n- 可以自动扫描最新 LangBot 后端日志，也可以扫描前端日志和 console 日志文件。\n- 可以用 `--since` 或 `--tail-lines` 把扫描范围限制到本次测试窗口。\n- 可以检测明显 Python/运行时错误和 secret 泄露风险。\n- 可以识别 case 声明的 success/failure patterns。\n- 可以识别 troubleshooting pattern，包括 `plugin-runtime-timeout` 和 `proxy-env-mismatch`。\n- 支持 case 级白名单。\n- 输出机器可读摘要。\n- 至少一个 `langbot-testing` case 使用它。\n\n当前 MVP 已覆盖自动发现 LangBot 后端日志、文件扫描、`--since`/`--tail-lines` 扫描窗口、\n基础错误检测、case success/failure signal、troubleshooting 匹配、secret 脱敏和 `--json`\n输出。仍待继续完善的是 live log 采集、更多规则、case 级 expected failure 的资产化和真实\nE2E report 样例。\n","isInternal":false,"tokens":1530,"sizeBytes":4684},{"name":"03-agent-browser-qa-principles.md","path":"skills/qa-agent-docs/qa-agent/03-agent-browser-qa-principles.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/qa-agent-docs/qa-agent/03-agent-browser-qa-principles.md","title":"Qa-agent Skill","category":"anthropic-skill","format":"markdown","content":"# Agent Browser QA Principles\n\nThis document fixes the direction of LangBot agent testing so the project does not drift into a backend API smoke-test framework.\n\n## Primary Goal\n\n`langbot-skills` should help an agent behave like a QA engineer using the product, not like a backend curl script.\n\nThe primary path is:\n\n```text\ndeveloper intent -> lbs test plan -> agent controls browser -> UI result + console + logs -> report/assets\n```\n\n## Rules\n\n1. Browser/UI interaction is the source of truth for product QA cases.\n2. A backend API or curl response is never enough to mark a UI case passed.\n3. API/curl/log checks are allowed as diagnostics after a UI path is attempted or when debugging environment readiness.\n4. A case passes only when the user-visible UI result is correct.\n5. The agent should inspect browser console/network output when available.\n6. If screenshot or vision capability is available, the agent should check for blank pages, overlap, hidden actions, broken layout, and error toasts.\n7. If no visual model is available, use DOM/accessibility snapshots and console output instead.\n8. New stable UI paths should be added as `cases/*.yaml`.\n9. New recurring failure modes should be added as `troubleshooting/*.yaml`.\n10. Secrets, tokens, API keys, and localStorage token values must never be printed.\n\n## Command Semantics\n\n`lbs` manages assets and produces plans. It does not replace the agent's browser-control ability.\n\n```bash\nbin/lbs test plan pipeline-debug-chat\n```\n\nThis command outputs:\n\n- environment variables to use\n- required skills\n- browser steps\n- UI/console/visual/log checks\n- diagnostic options\n- related troubleshooting patterns\n- report template\n\nThe active agent then executes the plan with Computer Use, Playwright MCP, or another available browser-control tool.\n\n## Diagnostics\n\nDiagnostics can include:\n\n- `bin/lbs env doctor`\n- browser console/network inspection\n- backend logs\n- targeted API/curl checks\n\nDiagnostics answer \"where did it fail?\" They do not replace \"did the user-visible UI work?\"\n","isInternal":false,"tokens":433,"sizeBytes":2042},{"name":"04-black-box-e2e-roadmap.md","path":"skills/qa-agent-docs/qa-agent/04-black-box-e2e-roadmap.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/qa-agent-docs/qa-agent/04-black-box-e2e-roadmap.md","title":"Qa-agent Skill","category":"anthropic-skill","format":"markdown","content":"# 黑盒 E2E QA 路线图\n\n## 定位\n\nLangBot 有大量外部依赖：模型供应商、plugin runtime、浏览器登录态、\nmarketplace/network、RAG engine、sandbox backend、平台适配器等。单测仍然有价值，\n但这个 QA 方向当前不优先解决 LangBot core 的单测覆盖率问题，因为重 mock 往往不能\n真实代表产品路径。\n\n`langbot-skills` 当前目标是让黑盒 E2E 测试变得可执行、可沉淀、可复用：\n\n```text\n开发者测试意图\n-> 复用或新增 case\n-> agent 通过浏览器执行\n-> UI + console + network + log 证据\n-> report\n-> 反哺 case / troubleshooting\n```\n\n这是面向开发者的 QA 资产库。开发者可以让 agent 测一个 feature；如果路径稳定，\n就把路径正规化为 case，让下一个开发者或 QA agent 继续复用。\n\n## 非目标\n\n- 这一阶段不优先建设 LangBot core 单测覆盖率。\n- 不把 API/curl 作为 WebUI 行为的通过标准。\n- 不要求每个 case 都能进 CI。\n- 不在 report 和日志守卫有用之前急着做完整 browser runner。\n- 不把外部 provider、OAuth、marketplace 抖动直接判成产品失败，除非证据明确。\n\n## 当前状态\n\n仓库已经具备第一层基础设施：\n\n- `skills/.env` 和 `skills/.env.local` 管理测试环境；\n- `langbot-env-setup`、`langbot-testing`、`langbot-plugin-dev` 等 skill；\n- `skills/langbot-testing/cases` 下的结构化 case；\n- `skills/langbot-testing/troubleshooting` 下的结构化故障资产；\n- RAG、多模态、plugin、MCP 等 fixture；\n- `bin/lbs validate`、`bin/lbs index`、`bin/lbs case`、`bin/lbs trouble`、\n `bin/lbs test plan`、`bin/lbs test start`、`bin/lbs test report`。\n\n所以当前已经不是“先把路径写进 Markdown”的阶段，而是进入“让每次运行有证据、\n有报告、能沉淀”的阶段。\n\n## 测试模型\n\nUI case 只有在用户可见行为正确时才能通过。辅助证据必须解释同一次运行。\n\n通过一个 UI case 的最低证据：\n\n- 用户可见的成功信号，例如 bot 回复、provider 保存成功、文件上传完成、plugin 页面渲染；\n- 没有意外 browser console error；\n- 相关时间窗口内没有意外后端/runtime 错误；\n- 有截图、DOM snapshot 或同等视觉/结构证据，如果当前 agent 能获取；\n- API/curl 只在解释同一条 UI 路径时作为诊断证据。\n\n失败报告需要保留足够信息，让开发者能复现或分流：\n\n- case id 和实际测试 URL；\n- 使用的 browser path；\n- 最后可见 UI 状态；\n- console/network 症状；\n- 相关后端/前端日志；\n- 匹配到的 troubleshooting id；\n- 这是产品失败、环境问题、外部依赖抖动，还是证据不足。\n\n## 结果词汇\n\n统一使用这些结果：\n\n- `pass`：UI 行为正确，辅助证据干净。\n- `fail`：UI 行为错误，或同一次运行的 console/log 出现意外产品错误。\n- `blocked`：缺登录、缺 provider credentials、服务未启动等原因导致目标路径没有跑起来。\n- `env_issue`：失败在目标行为之外，例如 proxy、OAuth、provider quota、marketplace outage、\n 本地服务启动问题。\n- `flaky`：同一环境下结果不稳定，进入门禁前需要先稳定。\n\n做 merge/release 判断时，`env_issue` 和 `blocked` 不能算产品通过。\n\n## 路线图\n\n### Phase 0：对齐文档\n\n目标：明确当前黑盒 E2E 方向。\n\n交付物：\n\n- `docs/qa-agent/README.md` 文档状态导航；\n- 本路线图；\n- 给旧规划文档加状态说明。\n\n完成标准：\n\n- 新贡献者不用通读所有旧文档，也能知道当前重点。\n\n### Phase 1：Test Report MVP\n\n状态：已有第一版。\n\n目标：让每次 agent browser 测试都有一致报告格式，即使 browser 执行还没自动化。\n\n建议命令：\n\n```bash\nbin/lbs test start \nbin/lbs test report --output reports/-.md\n```\n\nMVP 行为：\n\n- 读取 case 和关联 troubleshooting；\n- 生成 Markdown report 模板；\n- 生成 run handoff，固定本次测试的 start timestamp 和推荐 report command；\n- 写入脱敏后的环境摘要；\n- 提供 `pass/fail/blocked/env_issue/flaky` 结果选项；\n- 包含 UI result、console errors、network symptoms、logs、screenshots、\n diagnostics、matched troubleshooting、assets to update 等 section；\n- 支持 `--json`，输出机器可读报告。\n\n第一版已经是 report generator，不急着做自动判定。先把 evidence 收集格式统一起来，\n再做自动化更稳。\n\n完成标准：\n\n- agent 可以先跑 `lbs test start `，用它给出的时间窗口执行浏览器路径，\n 然后按固定格式填写 report，不需要每次重新发明报告结构。\n\n### Phase 2：日志守卫 MVP\n\n状态：已有第一版文件扫描。\n\n目标：捕获 UI 不一定明显展示的 runtime 问题。\n\n日志守卫应集成进 `lbs test report`，不要发展成独立后端 API 测试框架。\n\n建议命令形态：\n\n```bash\nbin/lbs test report \\\n --backend-log /path/to/backend.log \\\n --frontend-log /path/to/frontend.log \\\n --console-log /path/to/console.log \\\n --evidence-dir reports/evidence/ \\\n --since \"2026-05-21T10:30:00+08:00\" \\\n --tail-lines 2000 \\\n --output reports/-.md\n```\n\nMVP 行为：\n\n- 默认从 `LANGBOT_REPO/data/logs/` 扫描最新 `langbot-*.log`；\n- 支持 agent 显式提供 backend、frontend、console 日志文件；\n- 支持读取 evidence 目录下的 `automation-result.json`，把浏览器自动化脚本结论纳入报告；\n- 支持 `lbs test result` 为人工/agent browser 运行写入标准 `result.json`，供 suite 聚合；\n- 支持 `--since` 和 `--tail-lines`，避免历史日志污染本次报告；\n- 检测默认非预期模式，例如 `Traceback`、未 await coroutine、unclosed client/connector、\n `KeyError`、`TypeError`、`AttributeError`、明显 secret 泄露；\n- 匹配 case 声明的 `success_patterns` 和 `failure_patterns`；\n- 匹配已知 troubleshooting，先支持 `plugin-runtime-timeout` 和 `proxy-env-mismatch`；\n- 只有 case 明确声明时，才允许 expected failure；\n- 将发现分类为 fail、warning、matched troubleshooting、ignored expected issue；\n- 永远不打印 secret 值。\n\n完成标准：\n\n- 至少 `pipeline-debug-chat` 能生成包含日志摘要和 troubleshooting 匹配结果的 report。\n\n### Phase 3：Case 元数据加固\n\n状态：已有第一版。\n\n目标：让 case 更容易选择、执行和晋级。\n\n字段逐步补充，保持向后兼容：\n\n```yaml\npriority: p0 | p1 | p2\nrisk: low | medium | high\nci_eligible: false\npreconditions:\n - \"Authenticated browser profile is available.\"\nsetup:\n - \"Start LangBot backend and frontend.\"\ncleanup:\n - \"Remove temporary provider, plugin, or knowledge base if created.\"\nexpected_failures: []\nsuccess_patterns:\n - \"Conversation(0) Streaming completed\"\nfailure_patterns:\n - \"Action invoke_llm_stream call timed out\"\nevidence:\n required:\n - ui\n - console\n - backend_log\n```\n\n当前实现采用扁平字段 `evidence_required`，避免轻量 YAML 解析器在 case 文件里承载嵌套结构。\n`bin/lbs validate` 会校验 `priority`、`risk`、`ci_eligible`、`evidence_required`、\n`automation` 脚本路径、case 关联 skill 和 troubleshooting 交叉引用。`bin/lbs case list`\n支持 `--json`、`--type`、`--area`、`--tag`、`--priority`、`--risk`、`--automation`、`--ci`\n、`--ready` 和 `--machine-ready` 过滤，方便 agent 快速选择测试集。\n`env_any` 和 `automation_env_any` 用于表达 URL-or-name 这类 one-of 输入，避免把可替代变量误判为全部必填。\n\n当前也有 `skills//suites/*.yaml` 和 `bin/lbs suite plan `，用于组织常跑测试集，\n例如 `core-smoke`、`local-agent-gate` 和\n`agent-runner-release-gate`。发布门禁使用 `agent-runner-release-preflight`\n先分类配置 blockers 和 runtime env issues，再运行较重的浏览器 Debug Chat case。\n依赖 fixture 的 case 可以在浏览器执行前先跑 `bin/lbs fixture check`，检查\n`fixtures/fixtures.json` 登记的 deterministic 文件、plugin 包和本地测试 server 是否存在。\n`bin/lbs suite start ` 会生成 suite run id、suite evidence root、per-case evidence 目录、\n`suite-start.json`/`suite-start.md` handoff 文件和 per-case evidence 命令；\n浏览器自动化脚本会写入 `automation-result.json`，供 `bin/lbs test report` 展示原始自动化结论；\n`bin/lbs test result ` 会在人工/agent browser case 完成后写入最终 `result.json`；\n`bin/lbs suite report --evidence-dir ` 会聚合各 case 的 `result.json`，并且\n不会把缺少 required evidence 的 `pass` 当作 suite 通过。\nRunner 专用 Debug Chat case 通过 `automation_pipeline_url_env` 和\n`automation_pipeline_name_env` 绑定专用 pipeline 变量，避免 local-agent、Codex 或\nClaude Code case 误用通用 `LANGBOT_PIPELINE_URL` 后产生假阳性。\nDebug Chat case 还可以通过 `automation_stream_output` 固定流式或非流式发送路径。\n多模态 Debug Chat case 可以通过 `automation_image_base64_fixture` 复用 deterministic 图片 fixture。\n`test plan` 和 `suite plan` 会输出 readiness，让 agent 在执行浏览器前就看到缺失的 env、\n自动化变量、fixture，以及需要人工确认的 `manual_check` 前置条件。\n\n完成标准：\n\n- `lbs case list` 或后续 filter 能回答“smoke 跑哪些”、“哪些适合 CI”、\n “哪些需要真实 provider credentials”。\n\n### Phase 4：开发者沉淀流程\n\n目标：开发者让 agent 测新 feature 后，稳定路径不会丢在聊天记录里。\n\n流程：\n\n1. 开发者要求 agent 通过浏览器测试某个 feature。\n2. agent 先按 UI 主路径探索。\n3. agent 用 `lbs test start` 固定运行窗口，再用 `lbs test report` 写报告。\n4. 如果路径稳定，agent 新增或更新 case。\n5. 如果出现可复用故障，agent 新增或更新 troubleshooting。\n6. agent 跑 `bin/lbs validate` 和 `bin/lbs index`。\n\n完成标准：\n\n- feature QA 的结果能进入资产库，而不是只留在一次对话里。\n\n### Phase 5：选择性浏览器自动化\n\n状态：已有第一版 `test run` 入口和两个 Playwright 脚本。\n\n目标：只自动化少量稳定、值得重复跑的黑盒路径。\n\n建议顺序：\n\n1. `webui-login-state`\n2. `pipeline-debug-chat`\n3. `local-agent-basic-debug-chat`\n4. `local-agent-rag-debug-chat`\n5. 一个基于 deterministic fixture 的 plugin 或 MCP smoke path\n\n执行策略：\n\n- 继续把 Computer Use 或 Playwright MCP 作为默认交互路径；\n- 只给稳定、确定性的路径补直接 Playwright script；\n- 保存 screenshots、console logs、trace/video；\n- flaky 或强依赖真实 credentials 的 provider case 暂时不要进 CI。\n\n当前已经绑定：\n\n- `webui-login-state` -> `scripts/e2e/webui-login-state.mjs`\n- `pipeline-debug-chat` -> `scripts/e2e/pipeline-debug-chat.mjs`\n\n第一版自动化先产出 `reports/evidence//` 下的 console、network、screenshot 和\nresult JSON。真实执行后仍要用 `lbs test report --since ... --console-log ...` 做日志守卫和\n最终报告。开发期间可以先用 `bin/lbs test run --dry-run` 检查命令和 evidence 路径。\nDebug Chat 类脚本应复用 `scripts/e2e/lib/debug-chat.mjs`，避免重复实现 visible response leaf\n判断和已知失败信号分类。\n\n完成标准：\n\n- 小规模 smoke subset 可以不靠人工决定每一步点击；更大的资产库仍然服务于人工/agent\n 驱动的探索式 E2E。\n\n## 下一批动工切片\n\n在做 browser runner 之前，继续做这些：\n\n1. 等 LangBot 当前开发状态稳定后，用一次真实 `pipeline-debug-chat` 跑通\n `test start -> test run -> test report -> test result -> suite report`，产出 sample report。\n2. 只给 smoke/local-agent 首批 case 补必要元数据。\n3. 继续补日志守卫规则，尤其是 WebSocket、plugin runtime、provider streaming、前端\n chunk/rendering failure。\n4. 约定 report 产物目录、截图和 console/network 导出的命名方式。\n5. 再评估是否开始给 `webui-login-state` 和 `pipeline-debug-chat` 做直接 Playwright\n 自动化。\n\n这样 infra 会立刻有用，同时保留后续自动化 browser execution 的空间。\n","isInternal":false,"tokens":3915,"sizeBytes":12425},{"name":"README.md","path":"skills/qa-agent-docs/qa-agent/README.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/qa-agent-docs/qa-agent/README.md","title":"Qa-agent Skill","category":"anthropic-skill","format":"markdown","content":"# LangBot QA Agent 文档导航\n\n这个目录记录 `langbot-skills` 当前的 QA 方向和后续建设顺序。\n\n## 当前判断\n\n当前重点是 LangBot 的黑盒 E2E QA，不是 LangBot core 的单测覆盖率建设。\n\n`langbot-skills` 要帮助开发者和 QA agent 做接近人工测试的 WebUI 验证：\n\n- 打开真实 LangBot WebUI；\n- 按用户路径点击和输入；\n- 检查用户可见的 UI 结果；\n- 查看 console、network、截图、后端和前端日志；\n- 输出可复用的测试报告；\n- 把稳定 feature 路径沉淀为 case；\n- 把重复故障沉淀为 troubleshooting。\n\nAPI 和 curl 只做诊断。它们可以解释失败原因，但不能让一个 UI case 通过。\n\n## 文档状态\n\n| 文档 | 状态 | 用途 |\n| --- | --- | --- |\n| `04-black-box-e2e-roadmap.md` | 当前主路线图 | 决定下一步建设什么。 |\n| `03-agent-browser-qa-principles.md` | 当前原则文档 | 定义 browser-first QA 的通过标准。 |\n| `02-log-guard-plan.md` | 当前活跃设计 | 设计 `lbs test report` 里的日志守卫。 |\n| `../user-guide.md` | 当前使用手册 | 开发者日常使用。 |\n| `00-technology-options.md` | 背景文档 | 选择 Computer Use、Playwright MCP 或未来直接 Playwright。 |\n| `01-qa-agent-harness-plan.md` | 历史规划，部分过时 | 解释最初分层和目录设计；使用前先看状态说明。 |\n\n## 已过时的点\n\n`01-qa-agent-harness-plan.md` 还保留早期规划状态。现在结构化 cases、\n结构化 troubleshooting、`validate`、`index`、`lbs test plan` 都已经落地。\n\n已经补上第一版 `lbs test start`、`lbs test run`、`lbs test report` 和日志守卫文件扫描。\n`webui-login-state`、`pipeline-debug-chat` 已经绑定直接 Playwright 自动化脚本。后续重点是：\n\n- 报告 evidence 字段继续打磨；\n- case success/failure signal 和日志守卫规则继续补充；\n- 报告产物和 evidence 约定；\n- 等 LangBot 当前开发状态稳定后跑真实 sample report。\n\n不要再把旧阶段列表当成当前 source of truth。后续排序以\n`04-black-box-e2e-roadmap.md` 为准。\n","isInternal":false,"tokens":723,"sizeBytes":2122},{"name":"user-guide.md","path":"skills/qa-agent-docs/user-guide.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/qa-agent-docs/user-guide.md","title":"Qa-agent-docs Skill","category":"anthropic-skill","format":"markdown","content":"# LangBot Skills 用户使用手册\n\n## 这个仓库解决什么\n\n`langbot-skills` 是给 agent 使用的 LangBot 测试资产库。开发者 clone 后，可以让 Codex、Claude Code、Computer Use 或 Playwright MCP 复用已有环境配置、测试路径和故障知识，像 QA 一样操作 LangBot WebUI。\n\n核心目标：\n\n- 不让下一个 agent 窗口从头探索登录、模型配置、pipeline 调试。\n- 把稳定 UI 测试路径沉淀为 case。\n- 把常见故障沉淀为 troubleshooting。\n- 让 agent 优先通过浏览器点击验证产品行为。\n- API/curl/log 只作为诊断手段，不作为 UI case 通过标准。\n\n## 快速开始\n\n1. Clone 仓库。\n\n2. 检查本地默认变量：\n\n```bash\nbin/lbs env show\n```\n\n默认变量在：\n\n```text\nskills/.env\n```\n\n本机专用覆盖写到：\n\n```text\nskills/.env.local\n```\n\n它会覆盖 `skills/.env` 中的同名变量，并且不应该提交。\n`skills/.env` 是共享默认值，不应写入本机绝对路径、浏览器 profile、provider key 或其他凭据。\n新机器建议从模板开始：\n\n```bash\ncp skills/.env.example skills/.env.local\n```\n\n常用变量：\n\n```text\nLANGBOT_FRONTEND_URL\nLANGBOT_BACKEND_URL\nLANGBOT_DEV_FRONTEND_URL\nLANGBOT_REPO\nLANGBOT_WEB_REPO\nLANGBOT_BROWSER_PROFILE\n```\n\n3. 检查环境是否就绪：\n\n```bash\nbin/lbs env doctor\nbin/lbs fixture check\n```\n\n`env doctor` 会检查 URL、路径、代理变量等。代理变量是可选项；只有大小写代理变量互相冲突时才会报错。失败不一定代表仓库坏了，通常说明本地 LangBot 没启动、代理不一致或浏览器 profile 不存在。\n`fixture check` 会检查仓库内测试 fixture 是否存在，例如 MCP stdio server、RAG 文档、多模态图片、qa-plugin-smoke 包和 QA Runner 包。它也会校验 `.lbpkg` 是 zip 包，并检查 QA Runner fixture 的入口文件未漂移。\n\n4. 查看已有测试 case：\n\n```bash\nbin/lbs case list\nbin/lbs case list --json --priority p0 --automation\nbin/lbs case list --ready\nbin/lbs case list --machine-ready\nbin/lbs suite list\nbin/lbs suite plan core-smoke\nbin/lbs suite plan agent-runner-release-gate\nbin/lbs suite start core-smoke\nbin/lbs suite start core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\n```\n\n`case list` 支持按 `--type`、`--area`、`--tag`、`--priority`、`--risk`、`--automation`\n、`--ci`、`--ready` 和 `--machine-ready` 过滤。`--ready` 只显示没有缺机器输入且没有人工前置条件的 case；\n`--machine-ready` 过滤掉缺机器输入的 case，但保留 `manual-check`，表示执行前还要确认前置条件。需要交给 agent 自动选择测试集时，优先使用 `--json`，\n其中包含 `priority`、`risk`、`ci_eligible`、`automation`、`evidence_required` 以及\nenv/automation/fixture/manual readiness。\nCase metadata 中的 `env` 和 `automation_env` 表示全部必填；URL 或 name 这类二选一输入会放在\n`env_any` 或 `automation_env_any`，readiness 只要求组合里至少一个变量有值。\n\n如果要跑一组已沉淀的测试路径，优先使用 suite。Suite 位于 `skills//suites/*.yaml`，\n只负责组织 case，不改变 UI/browser 作为通过标准的原则。\n`suite plan` 会聚合 readiness：缺环境变量、缺自动化变量、缺 fixture 或需要\n`manual_check` 时，会在执行前标出受影响的 case。`manual_check` 不是产品通过，\n它表示机器配置已满足但 agent 必须先确认 case 里的 `preconditions` 或 `setup`。\nRunner externalization 发布判断使用 `agent-runner-release-gate`。先跑\n`agent-runner-release-preflight`，把缺 pipeline、runner id 错误、插件未安装这类\n`blocked`，以及 provider key、Box、插件运行时这类 `env_issue` 分开，再执行较重的\n浏览器 Debug Chat case。\n\n5. 生成 agent 执行计划：\n\n```bash\nbin/lbs test plan pipeline-debug-chat\n```\n\n然后把计划交给当前 agent 执行。agent 应使用 Computer Use、Playwright MCP 或其他浏览器控制能力去操作 UI。\n`test plan` 中的 Environment、Automation Readiness、Fixture Readiness 和 Manual\nReadiness 是执行前门禁；如果 readiness 缺失，应先补配置或将本次 case 标记为\n`blocked`。如果状态是 `manual_check`，先确认 `preconditions` 和 `setup`，再开始 UI\n执行。不要把后续 curl/API 诊断当成 UI case 通过。\n\n## 推荐使用方式\n\n### 冒烟测试\n\n你可以直接对 agent 说：\n\n```text\n帮我跑一下 LangBot 新前端 smoke test。\n```\n\nagent 应该：\n\n- 读 `skills/.env`\n- 优先查看 `bin/lbs suite plan core-smoke`，或查找 `type: smoke` 的 cases\n- 生成 test plan\n- 用浏览器执行 UI 操作\n- 检查 console、截图、后端日志\n- 输出简短 QA 报告\n\n### Runner Externalization 发布门禁\n\n你可以直接对 agent 说：\n\n```text\n按 agent-runner release gate 跑完整矩阵，先做 preflight，再跑浏览器 case，并把 blocked/env_issue/fail 分开。\n```\n\nagent 应该先查看 `skills/langbot-testing/references/agent-runner-release-gate.md`，\n再执行：\n\n```bash\nbin/lbs test recommend\nbin/lbs suite plan agent-runner-release-gate\nbin/lbs test run agent-runner-release-preflight --dry-run\nbin/lbs suite start agent-runner-release-gate --run-id agent-runner-release-local --evidence-dir reports/evidence/agent-runner-release-local\n```\n\n`test recommend` 输出的 run 命令默认带 `--dry-run`；确认 readiness 和 `manual_check` 前置条件后，再去掉 `--dry-run` 执行。\n\n完成所有 case 后，用：\n\n```bash\nbin/lbs suite report agent-runner-release-gate --evidence-dir reports/evidence/agent-runner-release-local\n```\n\n没有最终 `result.json`、缺 required evidence、或把 `blocked`/`env_issue` 当成 pass，\n都不能算发布门禁通过。\n\n### 新 Feature 测试\n\n你可以说：\n\n```text\n我改了 provider 页面，帮我测一下 DeepSeek provider 添加、测试、绑定 pipeline 是否正常。\n```\n\nagent 应该：\n\n- 查找相关 case 和 reference\n- 如果没有稳定路径，先探索 UI\n- 用浏览器执行真实交互\n- 失败时用日志/API 辅助定位\n- 稳定后新增或更新 case/reference\n- 新故障沉淀为 troubleshooting\n\n### 定点排错\n\n你可以说：\n\n```text\nDebug Chat 点了没回复，帮我查是前端问题还是后端问题。\n```\n\nagent 应该：\n\n- 先通过 UI 复现\n- 看 console/network\n- 看后端日志\n- 必要时用 API/curl 做诊断\n- 匹配 troubleshooting\n- 给出修复建议或直接修复\n\n## 重要原则\n\n这些原则固定在：\n\n```text\ndocs/qa-agent/03-agent-browser-qa-principles.md\n```\n\n简化版：\n\n- UI/browser 是测试主路径。\n- API/curl/log 只做诊断。\n- 后端接口成功不等于 UI case 通过。\n- case 通过必须以用户可见 UI 结果为准。\n- 有视觉能力时应检查截图。\n- 没有视觉能力时用 DOM/accessibility snapshot 和 console。\n- 不要打印 token、API key、OAuth secret 或 localStorage token 值。\n\n## 规划文档\n\n如果要判断下一步建设什么，先看：\n\n```text\ndocs/qa-agent/README.md\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n`01-qa-agent-harness-plan.md` 是早期规划，部分内容已经被当前实现和路线图替代。\n\n## 常用命令\n\n### 环境\n\n```bash\nbin/lbs env show\nbin/lbs env show --json\nbin/lbs env doctor\nbin/lbs fixture list\nbin/lbs fixture check\nbin/lbs fixture check --json\n```\n\n`env show` 和 `env doctor` 默认会对 token、API key、password、secret 以及 URL basic auth\n做脱敏。不要把 `.env.local` 里的原始凭据复制进测试报告。\n\n### Skill 和索引\n\n```bash\nbin/lbs list\nbin/lbs validate\nbin/lbs index --check\nbin/lbs index\n```\n\n### Case\n\n```bash\nbin/lbs case list\nbin/lbs case list --type smoke\nbin/lbs case list --json --priority p1 --tag local-agent\nbin/lbs case list --ready\nbin/lbs case list --machine-ready\nbin/lbs case show pipeline-debug-chat\nbin/lbs case new my-feature --title \"My Feature Works\"\n```\n\n### Suite\n\n```bash\nbin/lbs suite list\nbin/lbs suite list --json --priority p1\nbin/lbs suite show local-agent-gate\nbin/lbs suite plan core-smoke\nbin/lbs suite plan local-agent-gate --json\nbin/lbs suite start core-smoke\nbin/lbs suite start core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\nbin/lbs suite run core-smoke --dry-run --json\nbin/lbs suite run core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\nbin/lbs suite start core-smoke --json\nbin/lbs suite report core-smoke --evidence-dir reports/evidence/\nbin/lbs suite report core-smoke --evidence-dir reports/evidence/ --json\nbin/lbs suite new my-feature-gate --title \"My Feature Gate\"\n```\n\n`suite start` 不直接控制浏览器。它生成统一 run id、suite evidence root、每个 case 的 evidence\n目录、`suite-start.json`/`suite-start.md` handoff 文件，以及每个 case 的 `test run`、`test report`\n和 `test result` 命令模板。需要固定路径时，使用 `--run-id` 和 `--evidence-dir`。\n`suite run --dry-run --json` 只预览 planned/skipped case，不创建 evidence，也不执行 automation。\n`suite run` 会顺序执行 suite 中已有 automation、机器 readiness 已满足且不需要 `manual_check` 的 case，并在最后聚合 `suite report`。\n缺 env、automation env 或 fixture 的 case 默认会跳过；确实要强制执行时，加 `--include-not-ready`。\n确认前置条件后，才用 `--include-manual-check` 执行这类 case。\n所有 case 执行完并写入最终 `result.json` 后，\n`suite report` 会读取各 case evidence 目录并汇总为 `pass`、`fail`、`blocked`、`env_issue`、\n`flaky`、`incomplete` 等状态。`pass` 必须声明已经收集 case 的全部 required evidence；\n否则 suite 会保持 `incomplete`，避免把缺证据的运行误判成通过。\n\n### Test Plan\n\n```bash\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test plan pipeline-debug-chat --json\n```\n\n### Test Start\n\n```bash\nbin/lbs test start pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat --json\n```\n\n`test start` 用于 agent 开始一次浏览器测试前记录 run id、开始时间和推荐 report 命令。\n它会把 `--since \"\"` 写进后续报告命令，减少历史日志污染本次判断。\n如果 case 绑定了自动化脚本，输出里也会包含 `test run` 命令和 evidence 目录。\n\n### Test Automation\n\n```bash\nbin/lbs test run webui-login-state --dry-run\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test run webui-login-state --run-id login-smoke --output reports/evidence/login-smoke\nbin/lbs test run pipeline-debug-chat --run-id pipeline-smoke --output reports/evidence/pipeline-smoke\n```\n\n查看当前所有带自动化脚本的 case：\n\n```bash\nbin/lbs case list --automation\nbin/lbs case list --json --automation\n```\n\n当前自动化覆盖包括登录态、通用 Pipeline Debug Chat、local-agent runner 的基础回复、\nPromptPreProcessing、RAG、plugin tool、MCP stdio tool、非流式、多模态和 RAG+多模态路径。\n不要在文档里手工维护静态 case 清单；以 `case list --automation` 和 suite 定义为准。\n\n自动化脚本位于 `scripts/e2e/`。它们会保存：\n\n- `console.log`\n- `network.log`\n- `screenshot.png`\n- `automation-result.json`\n\n新增 Debug Chat 类自动化时，优先复用 `scripts/e2e/lib/debug-chat.mjs` 中的 pipeline 打开、\nprompt 发送、visible response leaf 判断和失败信号分类，不要在新脚本里复制 DOM 扫描逻辑。\n\n脚本需要本地安装 Playwright 后才能真正执行：\n\n```bash\nnpm install\nnpx playwright install chromium\n```\n\n`pipeline-debug-chat` 通用自动化建议配置 `LANGBOT_PIPELINE_URL`。如果没有 direct URL，\n脚本会尝试通过 `LANGBOT_PIPELINE_NAME` 从 Pipelines 页面进入目标 pipeline。两者都没有时，\n该自动化会返回 `blocked`，不会伪造通过。\n\nRunner 专用 case 不应复用通用 pipeline 变量。Local Agent、Codex Runner 和\nClaude Code Runner 这类 case 会通过 `automation_pipeline_url_env` /\n`automation_pipeline_name_env` 映射到 case-specific env，例如\n`LANGBOT_LOCAL_AGENT_PIPELINE_URL`。这些 case 如果缺少专用变量，会返回 `blocked`，\n不会退回到 `LANGBOT_PIPELINE_URL`，避免跑错 pipeline 后产生假阳性。\n如果 case 声明了 `setup_automation`，只有 `bin/lbs test run ` 的真实执行路径会先运行这些 setup；\n`test plan`、`suite plan`、`case list` 和 `--dry-run` 只展示它们，不会修改本地环境。\nsetup 可以是 `case:` 或仓库内 `node:scripts/... --flag`，每一步证据会写到主 evidence 目录下的\n`setup/` 子目录。setup 失败时主 automation 不会继续执行；setup 写入 `.env.local` 后，主 automation\n会重新读取环境。用 `setup_provides_env` 声明 setup 会生成的变量，可以让 readiness 正确显示机器可准备状态。\n如果 Debug Chat case 需要固定流式/非流式路径，可以在 case 中设置\n`automation_stream_output: \"1\"` 或 `\"0\"`，脚本会在发送消息前切换 Debug Chat 的 Stream 控件。\n如果 case 需要上传图片，可以设置 `automation_image_base64_fixture` 指向仓库内的 base64 PNG fixture，\n脚本会在 evidence 目录写出临时 PNG 并通过 Debug Chat 上传控件发送。\n`bin/lbs test plan --json` 和 `bin/lbs suite plan --json`\n都会显示这些专用变量是否已配置。\n\n### Test Report 和日志守卫\n\n```bash\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report pipeline-debug-chat --output reports/pipeline-debug-chat.md\nbin/lbs test report pipeline-debug-chat \\\n --backend-log /path/to/backend.log \\\n --frontend-log /path/to/frontend.log \\\n --console-log /path/to/console.log\nbin/lbs test report pipeline-debug-chat --evidence-dir reports/evidence/pipeline-smoke\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log --json\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs test report pipeline-debug-chat --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\" --tail-lines 2000\n```\n\n`test report` 会生成报告模板，并默认从 `LANGBOT_REPO/data/logs/` 自动选择最新的\n`langbot-*.log` 作为 LangBot 后端日志扫描。也可以用 `--backend-log` 覆盖，或用\n`--no-auto-log` 只生成模板。\n\n如果提供 `--evidence-dir`，或 `--console-log` 指向 `reports/evidence//console.log`，\n报告会优先读取同目录的 `automation-result.json`，并展示自动化脚本的 `status`、`reason`、\n起止时间和目标 URL。\n\n日志守卫会扫描常见错误、secret 泄露风险、case 声明的 success/failure patterns，以及已知\ntroubleshooting pattern。它不控制浏览器，也不替代 UI 通过判断。`success_patterns`\n命中会作为通过证据写入 `success_signals`；声明了 success pattern 但本次扫描窗口没有命中，\n会给 warning；`failure_patterns` 命中会让本次日志守卫 fail。\n\n建议在执行浏览器 case 前记录当前时间，然后在报告阶段使用 `--since`。如果只想快速看\n最近日志，可以使用 `--tail-lines`。\n\n### Runtime Log Guard\n\n如果还没有进入某个具体 UI case，只是想观察 LangBot 后端日志，可以直接使用 `log`\n命令。它和 `test report` 使用同一套扫描器、secret 脱敏、troubleshooting pattern 和\ncase success/failure pattern。\n\n```bash\nbin/lbs log scan --tail-lines 300\nbin/lbs log scan --case pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs log scan --backend-log /path/to/langbot.log --json\nbin/lbs log scan --failure-pattern \"runner.tool_loop_error|Action invoke_llm_stream call timed out\" --strict\n```\n\n`log scan` 默认从 `LANGBOT_REPO/data/logs/` 自动选择最新的 `langbot-*.log`。传入\n`--case ` 后，会额外应用该 case 声明的 `success_patterns`、`failure_patterns`\n和 related troubleshooting。默认用于观察，返回码保持 0；加 `--strict` 后，`fail` 或\n`env_issue` 会返回非 0，适合脚本门禁。\n\n运行期观察可以用 `watch`：\n\n```bash\nbin/lbs log watch --case pipeline-debug-chat\nbin/lbs log watch --backend-log /path/to/langbot.log --interval-ms 500\nbin/lbs log watch --duration-ms 30000 --strict --json\n```\n\n`log watch` 默认从启动时的文件末尾开始，只观察新追加的日志；加 `--from-start` 可从文件开头扫。\n它会实时打印新命中的 findings 和 success signals。为了避免当前历史日志噪声影响观察，默认不因\n异常返回非 0；加 `--strict` 后，退出时如果看到 `fail` 或 `env_issue` 会返回非 0。\n\n给一次 QA 运行包日志窗口时，用 `guard start/stop`：\n\n```bash\nbin/lbs log guard start --run-id local-debug --case pipeline-debug-chat\n# 执行浏览器或手工测试\nbin/lbs log guard stop --run-id local-debug\n```\n\n`start` 会在 `reports/log-guards/.json` 记录起始时间、case 和当前后端日志路径；\n`stop` 会用 start/stop 时间作为扫描窗口，生成 `reports/log-guards/.md`，并默认按\nstrict guard 返回码处理。临时只想收集报告、不想让命令失败，可以加 `--no-strict`。\n\n当前 LangBot core 日志还不是完全结构化日志，runtime guard 主要依赖时间窗口和文本 pattern。\n已支持 ISO 时间戳和 LangBot 当前的 `[MM-DD HH:mm:ss.SSS]` 前缀；没有时间戳的连续行会跟随上一条\n带时间戳的日志块。如果后续 core 能提供稳定 request id、conversation id、plugin action id 或\nJSON log field，guard 可以从“时间窗口 + 文本匹配”升级为更精确的关联分析。\n\n### Test Result\n\n```bash\nbin/lbs test result pipeline-debug-chat \\\n --result pass \\\n --reason \"Debug Chat returned OK and the report log guard was clean.\" \\\n --evidence-dir reports/evidence/pipeline-smoke \\\n --started-at \"2026-05-21T10:30:00+08:00\" \\\n --evidence ui,screenshot,console,backend_log\n```\n\n`test result` 用于把一次人工/agent browser 运行的最终判断写成标准 `result.json`，\n供 `suite report` 聚合。它不会替代 UI 测试：如果写 `--result pass`，`--evidence`\n必须覆盖该 case 的 `evidence_required`，否则命令会失败。自动化脚本写\n`automation-result.json`；如果 case 还要求 backend log、API diagnostic 或 filesystem\nevidence，agent 需要在报告和诊断完成后再用 `test result` 写最终 `result.json`。\n\n### Troubleshooting\n\n```bash\nbin/lbs trouble list langbot-testing\nbin/lbs trouble show plugin-runtime-timeout\nbin/lbs trouble search runtime\nbin/lbs trouble add langbot-testing --title \"...\" --symptom \"...\" --cause \"...\" --fix \"...\"\n```\n\n## 目录说明\n\n```text\nskills/\n .env # 共享默认变量\n langbot-env-setup/ # 环境、浏览器、登录态、代理\n langbot-testing/ # WebUI / provider / pipeline 测试\nschemas/ # 结构化资产 schema\nsrc/ # lbs TypeScript 源码\nbin/ # lbs 入口\ndocs/ # 设计文档和用户手册\nAGENTS.md # agent 维护协议\n```\n\n## 添加一个新测试路径\n\n1. 先让 agent 通过浏览器探索并执行路径。\n2. 稳定后创建 case：\n\n```bash\nbin/lbs case new provider-xxx --title \"Provider XXX can be configured\" --area provider --type provider\n```\n\n3. 编辑生成的 `cases/*.yaml`，补充真实步骤、检查点和 troubleshooting。\n\n4. 校验：\n\n```bash\nbin/lbs validate\nbin/lbs index --check\nbin/lbs index\n```\n\n## 添加一个新故障\n\n```bash\nbin/lbs trouble add langbot-testing \\\n --title \"Plugin runtime actions time out\" \\\n --symptom \"Debug Chat shows Agent runner temporarily unavailable\" \\\n --cause \"Old plugin runtime survived backend restart\" \\\n --fix \"Stop old runtime processes and restart LangBot\"\n```\n\n然后编辑生成的 YAML，补充 `patterns`、`related_cases` 和验证方式。\n\n## 当前边界\n\n- `lbs test plan` 只生成测试计划，不直接控制浏览器。\n- `lbs test report` 生成报告，默认扫描最新 LangBot 后端日志；也可扫描显式提供的\n backend/frontend/console 日志文件。\n- 真正的 UI 操作由当前 agent 的浏览器能力执行。\n- `env doctor` 是 readiness check，不是产品测试。\n- `curl/API` 是诊断工具，不是主要测试路径。\n","isInternal":false,"tokens":6248,"sizeBytes":20489},{"name":"README.md","path":"skills/README.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/README.md","title":"Skills Skill","category":"anthropic-skill","format":"markdown","content":"# LangBot Skills\n\nThis directory is the **single source of truth** for LangBot's agent skills —\nreusable, on-demand instruction packs for AI agents (Claude Code, Codex, Cursor,\nand LangBot's own Local Agent) working with the LangBot ecosystem.\n\n> These skills were consolidated here from the former `langbot-app/langbot-skills`\n> repository (now archived). Documentation and the landing page link here; do not\n> re-copy skill content elsewhere — link to this directory instead.\n\n## Skill catalog\n\n| Skill | What it covers |\n| --- | --- |\n| [`langbot-dev`](skills/langbot-dev) | Core backend + web frontend development (Quart, Vite, API, migrations, MCP server) |\n| [`langbot-plugin-dev`](skills/langbot-plugin-dev) | Plugin SDK / component development, debugging, WebSocket testing |\n| [`langbot-deploy`](skills/langbot-deploy) | Docker / Compose / Kubernetes deployment, config.yaml, Box runtime, global API key |\n| [`langbot-testing`](skills/langbot-testing) | WebUI / e2e QA harness, cases, fixtures, troubleshooting (the `bin/lbs` CLI) |\n| [`langbot-env-setup`](skills/langbot-env-setup) | Local dev/test environment, browser access, OAuth, proxy, startup |\n| [`langbot-mcp-ops`](skills/langbot-mcp-ops) | Operating a LangBot instance through its MCP server (`/mcp`) |\n| [`langbot-space-ops`](skills/langbot-space-ops) | Browsing the LangBot Space marketplaces through the Space MCP server |\n| [`langbot-eba-adapter-dev`](skills/langbot-eba-adapter-dev) | Building platform adapters for the Event-Based Agents architecture |\n| [`langbot-skills-maintenance`](skills/langbot-skills-maintenance) | Adding, deduplicating, and auditing skills in this directory |\n\n`skills.index.json` is the machine-readable index (regenerate with `bin/lbs index`).\n\n## Quick start (for an AI agent)\n\n1. Read this README, `AGENTS.md`, and `docs/user-guide.md` to understand the layout.\n2. Read `skills/.env` for shared local defaults. On a new machine, copy\n `skills/.env.example` to `skills/.env.local` (gitignored) and override\n machine-specific values there. Never commit secrets.\n3. Pick the smallest relevant skill from the catalog above and follow its\n `SKILL.md`.\n\n## The `lbs` CLI\n\nThe testing assets ship with a small CLI (`bin/lbs`, Node >= 22.6). The\n`bin/lbs` wrapper is a generated local entrypoint; on a fresh checkout, run\n`npm run bootstrap` once if it is missing. `npm install` also regenerates it via\nthe `prepare` script.\n\n```bash\nnpm run bootstrap # create bin/lbs if missing\nbin/lbs validate # validate skills/cases/troubleshooting structure\nbin/lbs index # regenerate skills.index.json\nbin/lbs env show # inspect resolved env defaults (redacted)\nbin/lbs env doctor # diagnose local environment readiness\nbin/lbs case list --ready\nbin/lbs test plan \nbin/lbs suite plan langbot-debug-chat-load-gate\n```\n\n## Maintenance rule\n\nWhen the LangBot / LangBot Space **API or MCP server changes**, the\ncorresponding skill here MUST be updated in the same change. The MCP tool\nsurface, the API, and these skills are kept in lockstep — see each repo's\n`AGENTS.md`.\n","isInternal":false,"tokens":764,"sizeBytes":3097},{"name":"case.schema.json","path":"skills/schemas/case.schema.json","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/schemas/case.schema.json","title":"Schemas Skill","category":"anthropic-skill","format":"json","content":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/case.schema.json\",\n \"title\": \"LangBot Skill Test Case\",\n \"type\": \"object\",\n \"required\": [\n \"id\",\n \"title\",\n \"mode\",\n \"area\",\n \"type\",\n \"priority\",\n \"risk\",\n \"ci_eligible\",\n \"tags\",\n \"skills\",\n \"steps\",\n \"checks\",\n \"evidence_required\"\n ],\n \"allOf\": [\n {\n \"if\": {\n \"properties\": {\n \"mode\": { \"const\": \"agent-browser\" }\n }\n },\n \"then\": {\n \"required\": [\"env\"]\n }\n }\n ],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"mode\": {\n \"type\": \"string\",\n \"enum\": [\"agent-browser\", \"probe\"]\n },\n \"area\": {\n \"type\": \"string\"\n },\n \"type\": {\n \"type\": \"string\",\n \"enum\": [\n \"smoke\",\n \"regression\",\n \"feature\",\n \"provider\",\n \"exploratory\",\n \"contract\",\n \"performance\",\n \"reliability\",\n \"chaos\",\n \"security\"\n ]\n },\n \"priority\": {\n \"type\": \"string\",\n \"enum\": [\"p0\", \"p1\", \"p2\"]\n },\n \"risk\": {\n \"type\": \"string\",\n \"enum\": [\"low\", \"medium\", \"high\"]\n },\n \"ci_eligible\": {\n \"type\": \"boolean\"\n },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"skills\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"env_any\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*(\\\\|[A-Z][A-Z0-9_]*)+$\"\n }\n },\n \"steps\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"checks\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"evidence_required\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"enum\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"backend_log\",\n \"frontend_log\",\n \"api_diagnostic\",\n \"filesystem\",\n \"metrics\",\n \"trace\",\n \"profile\",\n \"resource_log\"\n ]\n },\n \"minItems\": 1\n },\n \"preconditions\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup_automation\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^(?:case:[a-z0-9][a-z0-9_-]*|node:scripts/[A-Za-z0-9_./-]+\\\\.(?:mjs|js|ts)(?:\\\\s+--[A-Za-z0-9][A-Za-z0-9_-]*(?:=[A-Za-z0-9_./:@-]+)?)*)$\"\n }\n },\n \"setup_provides_env\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n }\n },\n \"cleanup\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"diagnostics\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation\": {\n \"type\": \"string\"\n },\n \"automation_env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation_env_any\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*(\\\\|[A-Z][A-Z0-9_]*)+$\"\n }\n },\n \"automation_prompt\": {\n \"type\": \"string\"\n },\n \"automation_prompts_json\": {\n \"type\": \"string\"\n },\n \"automation_expected_text\": {\n \"type\": \"string\"\n },\n \"automation_response_timeout_ms\": {\n \"type\": \"string\"\n },\n \"automation_stream_output\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_image_base64_fixture\": {\n \"type\": \"string\"\n },\n \"automation_runner_config_patch_json\": {\n \"type\": \"string\"\n },\n \"automation_restore_runner_config\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_expected_runner_id\": {\n \"type\": \"string\"\n },\n \"automation_reset_debug_chat\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_session_type\": {\n \"type\": \"string\",\n \"enum\": [\"person\", \"group\"]\n },\n \"automation_debug_chat_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_max_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_requests\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_concurrency\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_timeout_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_first_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_max_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_error_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_ok_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_provider_fault_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_expected_prefix\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_prompt_template\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_stream\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_reset\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_fail_on_final_mismatch\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_require_success\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_provider_model_thresholds_json\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_pipeline_name\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_model_name\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fallback_model_names\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_response_text\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_first_token_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_chunk_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_chunk_count\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_first_n\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_every_n\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_models\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fault_status\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_after_first_chunk\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_fake_provider_fail_after_first_chunk_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_after_first_chunk_mode\": {\n \"type\": \"string\",\n \"enum\": [\"disconnect\", \"error_event\"]\n },\n \"automation_fake_provider_fail_after_first_chunk_models\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_dynamic_response\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_filesystem_checks_json\": {\n \"type\": \"string\"\n },\n \"metrics_thresholds_json\": {\n \"type\": \"string\"\n },\n \"load_profile_json\": {\n \"type\": \"string\"\n },\n \"fault_model_json\": {\n \"type\": \"string\"\n },\n \"automation_pipeline_url_env\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n },\n \"automation_pipeline_name_env\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n },\n \"success_patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"failure_patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"expected_failures\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"troubleshooting\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n}\n","frontmatter":{"$schema":"https://json-schema.org/draft/2020-12/schema","$id":"https://langbot.app/schemas/langbot-skills/case.schema.json","title":"LangBot Skill Test Case","type":"object","required":["id","title","mode","area","type","priority","risk","ci_eligible","tags","skills","steps","checks","evidence_required"],"allOf":[{"if":{"properties":{"mode":{"const":"agent-browser"}}},"then":{"required":["env"]}}],"additionalProperties":true,"properties":{"id":{"type":"string","pattern":"^[a-z0-9][a-z0-9_-]*$"},"title":{"type":"string"},"mode":{"type":"string","enum":["agent-browser","probe"]},"area":{"type":"string"},"type":{"type":"string","enum":["smoke","regression","feature","provider","exploratory","contract","performance","reliability","chaos","security"]},"priority":{"type":"string","enum":["p0","p1","p2"]},"risk":{"type":"string","enum":["low","medium","high"]},"ci_eligible":{"type":"boolean"},"tags":{"type":"array","items":{"type":"string"}},"skills":{"type":"array","items":{"type":"string"}},"env":{"type":"array","items":{"type":"string"}},"env_any":{"type":"array","items":{"type":"string","pattern":"^[A-Z][A-Z0-9_]*(\\|[A-Z][A-Z0-9_]*)+$"}},"steps":{"type":"array","items":{"type":"string"},"minItems":1},"checks":{"type":"array","items":{"type":"string"},"minItems":1},"evidence_required":{"type":"array","items":{"type":"string","enum":["ui","screenshot","console","network","backend_log","frontend_log","api_diagnostic","filesystem","metrics","trace","profile","resource_log"]},"minItems":1},"preconditions":{"type":"array","items":{"type":"string"}},"setup":{"type":"array","items":{"type":"string"}},"setup_automation":{"type":"array","items":{"type":"string","pattern":"^(?:case:[a-z0-9][a-z0-9_-]*|node:scripts/[A-Za-z0-9_./-]+\\.(?:mjs|js|ts)(?:\\s+--[A-Za-z0-9][A-Za-z0-9_-]*(?:=[A-Za-z0-9_./:@-]+)?)*)$"}},"setup_provides_env":{"type":"array","items":{"type":"string","pattern":"^[A-Z][A-Z0-9_]*$"}},"cleanup":{"type":"array","items":{"type":"string"}},"diagnostics":{"type":"array","items":{"type":"string"}},"automation":{"type":"string"},"automation_env":{"type":"array","items":{"type":"string"}},"automation_env_any":{"type":"array","items":{"type":"string","pattern":"^[A-Z][A-Z0-9_]*(\\|[A-Z][A-Z0-9_]*)+$"}},"automation_prompt":{"type":"string"},"automation_prompts_json":{"type":"string"},"automation_expected_text":{"type":"string"},"automation_response_timeout_ms":{"type":"string"},"automation_stream_output":{"type":"string","enum":["0","1","false","true"]},"automation_image_base64_fixture":{"type":"string"},"automation_runner_config_patch_json":{"type":"string"},"automation_restore_runner_config":{"type":"string","enum":["0","1","false","true"]},"automation_expected_runner_id":{"type":"string"},"automation_reset_debug_chat":{"type":"string","enum":["0","1","false","true"]},"automation_debug_chat_session_type":{"type":"string","enum":["person","group"]},"automation_debug_chat_response_p95_ms":{"type":"string"},"automation_debug_chat_max_error_rate":{"type":"string"},"automation_debug_chat_load_requests":{"type":"string"},"automation_debug_chat_load_concurrency":{"type":"string"},"automation_debug_chat_load_timeout_ms":{"type":"string"},"automation_debug_chat_load_response_p95_ms":{"type":"string"},"automation_debug_chat_load_first_response_p95_ms":{"type":"string"},"automation_debug_chat_load_max_error_rate":{"type":"string"},"automation_debug_chat_load_min_error_rate":{"type":"string"},"automation_debug_chat_load_min_error_count":{"type":"string"},"automation_debug_chat_load_min_ok_count":{"type":"string"},"automation_debug_chat_load_min_provider_fault_count":{"type":"string"},"automation_debug_chat_load_expected_prefix":{"type":"string"},"automation_debug_chat_load_prompt_template":{"type":"string"},"automation_debug_chat_load_stream":{"type":"string","enum":["0","1","false","true"]},"automation_debug_chat_load_reset":{"type":"string","enum":["0","1","false","true"]},"automation_debug_chat_load_fail_on_final_mismatch":{"type":"string","enum":["0","1","false","true"]},"automation_debug_chat_load_require_success":{"type":"string","enum":["0","1","false","true"]},"automation_debug_chat_load_provider_model_thresholds_json":{"type":"string"},"automation_fake_provider_pipeline_name":{"type":"string"},"automation_fake_provider_model_name":{"type":"string"},"automation_fake_provider_fallback_model_names":{"type":"string"},"automation_fake_provider_response_text":{"type":"string"},"automation_fake_provider_first_token_delay_ms":{"type":"string"},"automation_fake_provider_chunk_delay_ms":{"type":"string"},"automation_fake_provider_chunk_count":{"type":"string"},"automation_fake_provider_fail_first_n":{"type":"string"},"automation_fake_provider_fail_every_n":{"type":"string"},"automation_fake_provider_fail_models":{"type":"string"},"automation_fake_provider_fault_status":{"type":"string"},"automation_fake_provider_fail_after_first_chunk":{"type":"string","enum":["0","1","false","true"]},"automation_fake_provider_fail_after_first_chunk_delay_ms":{"type":"string"},"automation_fake_provider_fail_after_first_chunk_mode":{"type":"string","enum":["disconnect","error_event"]},"automation_fake_provider_fail_after_first_chunk_models":{"type":"string"},"automation_fake_provider_dynamic_response":{"type":"string","enum":["0","1","false","true"]},"automation_filesystem_checks_json":{"type":"string"},"metrics_thresholds_json":{"type":"string"},"load_profile_json":{"type":"string"},"fault_model_json":{"type":"string"},"automation_pipeline_url_env":{"type":"string","pattern":"^[A-Z][A-Z0-9_]*$"},"automation_pipeline_name_env":{"type":"string","pattern":"^[A-Z][A-Z0-9_]*$"},"success_patterns":{"type":"array","items":{"type":"string"}},"failure_patterns":{"type":"array","items":{"type":"string"}},"expected_failures":{"type":"array","items":{"type":"string"}},"troubleshooting":{"type":"array","items":{"type":"string"}}}},"isInternal":false,"tokens":2377,"sizeBytes":8380},{"name":"README.md","path":"skills/schemas/README.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/schemas/README.md","title":"Schemas Skill","category":"anthropic-skill","format":"markdown","content":"# Schemas\n\n这个目录存放 LangBot skills 结构化资产的 JSON Schema。\n\n它们不是测试脚本，也不会执行浏览器动作。它们的作用是定义 agent 和维护者后续新增资产时应该遵守的文件结构。\n\n## 文件说明\n\n- `skills//fixtures/fixtures.json`\n 不是 JSON Schema，但由 `bin/lbs validate` 校验。\n 它登记 deterministic fixture 文件、类型和关联 case，供 `bin/lbs fixture check` 做 readiness 检查。\n\n- `case.schema.json`\n 约束 `skills//cases/*.yaml` 的格式。\n Case 描述 agent-browser 或 probe QA 路径，包括前置条件、步骤、检查点、诊断手段和关联故障。\n\n- `suite.schema.json`\n 约束 `skills//suites/*.yaml` 的格式。\n Suite 只组织 case 集合，用于 smoke、regression 或 release gate 等测试入口。\n\n- `troubleshooting.schema.json`\n 约束 `skills//troubleshooting/*.yaml` 的格式。\n Troubleshooting 条目描述症状、日志/错误模式、可能原因、修复步骤和验证信号。\n\n- `skill-index.schema.json`\n 约束生成文件 `skills.index.json` 的格式。\n 这个索引用于让 agent 快速发现已有 skills、references、cases、suites 和 troubleshooting。\n\n- `reports/evidence//result.json`\n 不是 catalog schema，而是执行期最终裁定产物，由 `bin/lbs test result` 写入。\n `suite report` 读取其中的 `status`、`reason`、起止时间和 `evidence_collected`，\n 并用 `evidence_missing` 防止缺证据的 `pass` 被当作完整通过。\n\n- `reports/evidence//automation-result.json`\n 不是 catalog schema，而是浏览器自动化脚本的原始运行结论，供 `bin/lbs test report`\n 展示和推断日志扫描窗口。\n\n## 为什么需要 schemas\n\nSchemas 是基础设施护栏：\n\n- 防止 case、suite 和 troubleshooting 随着增长变得格式混乱\n- 让 `bin/lbs validate` 能发现缺字段和错误结构\n- 为未来编辑器提示和 CI 校验留接口\n- 帮助 agent 新增资产时知道应该写哪些字段\n\n## 当前校验方式\n\n`bin/lbs validate` 做轻量、schema 对齐的校验，不引入额外依赖。它会检查必填字段、\n枚举值、boolean 字段、重复列表项、automation 脚本存在性，以及 case、suite、skill、\ntroubleshooting 之间的交叉引用。这里的 schema 仍是格式契约；如果未来引入正式 JSON\nSchema validator，应继续保持这些本地交叉引用检查。\n\nCase 里的 `env` / `automation_env` 表示所有列出的变量都需要配置。遇到二选一输入时，\n使用 `env_any` / `automation_env_any`，每一项写成 `LANGBOT_PIPELINE_URL|LANGBOT_PIPELINE_NAME`\n这类 one-of 组合，避免 agent 因为只配置了 URL 或 name 其中之一而误判未就绪。\n`setup` 和 `preconditions` 是人工确认项，会让 readiness 进入 `manual_check`；\n`setup_automation` 是 `test run` 可以自动执行的准备步骤，配合 `setup_provides_env`\n声明它会生成的机器变量。\n","isInternal":false,"tokens":960,"sizeBytes":3006},{"name":"skill-index.schema.json","path":"skills/schemas/skill-index.schema.json","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/schemas/skill-index.schema.json","title":"Schemas Skill","category":"anthropic-skill","format":"json","content":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/skill-index.schema.json\",\n \"title\": \"LangBot Skills Index\",\n \"type\": \"object\",\n \"required\": [\"generated_by\", \"skills\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"generated_by\": {\n \"type\": \"string\"\n },\n \"skills\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\n \"directory\",\n \"name\",\n \"description\",\n \"references\",\n \"cases\",\n \"case_summaries\",\n \"suites\",\n \"suite_summaries\",\n \"fixtures\",\n \"troubleshooting\",\n \"troubleshooting_summaries\"\n ],\n \"additionalProperties\": false,\n \"properties\": {\n \"directory\": { \"type\": \"string\" },\n \"name\": { \"type\": \"string\" },\n \"description\": { \"type\": \"string\" },\n \"references\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"case_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"mode\", \"area\", \"type\", \"priority\", \"risk\", \"ci_eligible\", \"tags\", \"automation\", \"setup_automation\", \"setup_provides_env\", \"evidence_required\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"mode\": { \"type\": \"string\", \"enum\": [\"agent-browser\", \"probe\"] },\n \"area\": { \"type\": \"string\" },\n \"type\": { \"type\": \"string\" },\n \"priority\": { \"type\": \"string\" },\n \"risk\": { \"type\": \"string\" },\n \"ci_eligible\": { \"type\": \"boolean\" },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation\": { \"type\": \"string\" },\n \"setup_automation\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup_provides_env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"evidence_required\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"suites\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"suite_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"description\", \"type\", \"priority\", \"tags\", \"cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"description\": { \"type\": \"string\" },\n \"type\": { \"type\": \"string\" },\n \"priority\": { \"type\": \"string\" },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"fixtures\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"kind\", \"path\", \"related_cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"kind\": { \"type\": \"string\" },\n \"path\": { \"type\": \"string\" },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"troubleshooting\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"troubleshooting_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"category\", \"related_cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"category\": { \"type\": \"string\" },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n }\n }\n }\n }\n }\n}\n","frontmatter":{"$schema":"https://json-schema.org/draft/2020-12/schema","$id":"https://langbot.app/schemas/langbot-skills/skill-index.schema.json","title":"LangBot Skills Index","type":"object","required":["generated_by","skills"],"additionalProperties":false,"properties":{"generated_by":{"type":"string"},"skills":{"type":"array","items":{"type":"object","required":["directory","name","description","references","cases","case_summaries","suites","suite_summaries","fixtures","troubleshooting","troubleshooting_summaries"],"additionalProperties":false,"properties":{"directory":{"type":"string"},"name":{"type":"string"},"description":{"type":"string"},"references":{"type":"array","items":{"type":"string"}},"cases":{"type":"array","items":{"type":"string"}},"case_summaries":{"type":"array","items":{"type":"object","required":["id","title","mode","area","type","priority","risk","ci_eligible","tags","automation","setup_automation","setup_provides_env","evidence_required"],"additionalProperties":false,"properties":{"id":{"type":"string"},"title":{"type":"string"},"mode":{"type":"string","enum":["agent-browser","probe"]},"area":{"type":"string"},"type":{"type":"string"},"priority":{"type":"string"},"risk":{"type":"string"},"ci_eligible":{"type":"boolean"},"tags":{"type":"array","items":{"type":"string"}},"automation":{"type":"string"},"setup_automation":{"type":"array","items":{"type":"string"}},"setup_provides_env":{"type":"array","items":{"type":"string"}},"evidence_required":{"type":"array","items":{"type":"string"}}}}},"suites":{"type":"array","items":{"type":"string"}},"suite_summaries":{"type":"array","items":{"type":"object","required":["id","title","description","type","priority","tags","cases"],"additionalProperties":false,"properties":{"id":{"type":"string"},"title":{"type":"string"},"description":{"type":"string"},"type":{"type":"string"},"priority":{"type":"string"},"tags":{"type":"array","items":{"type":"string"}},"cases":{"type":"array","items":{"type":"string"}}}}},"fixtures":{"type":"array","items":{"type":"object","required":["id","title","kind","path","related_cases"],"additionalProperties":false,"properties":{"id":{"type":"string"},"title":{"type":"string"},"kind":{"type":"string"},"path":{"type":"string"},"related_cases":{"type":"array","items":{"type":"string"}}}}},"troubleshooting":{"type":"array","items":{"type":"string"}},"troubleshooting_summaries":{"type":"array","items":{"type":"object","required":["id","title","category","related_cases"],"additionalProperties":false,"properties":{"id":{"type":"string"},"title":{"type":"string"},"category":{"type":"string"},"related_cases":{"type":"array","items":{"type":"string"}}}}}}}}}},"isInternal":false,"tokens":1120,"sizeBytes":4865},{"name":"suite.schema.json","path":"skills/schemas/suite.schema.json","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/schemas/suite.schema.json","title":"Schemas Skill","category":"anthropic-skill","format":"json","content":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/suite.schema.json\",\n \"title\": \"LangBot Skill Test Suite\",\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"description\", \"type\", \"priority\", \"tags\", \"cases\"],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"description\": {\n \"type\": \"string\"\n },\n \"type\": {\n \"type\": \"string\",\n \"enum\": [\n \"smoke\",\n \"regression\",\n \"release_gate\",\n \"exploratory\",\n \"contract\",\n \"performance\",\n \"reliability\",\n \"chaos\",\n \"security\"\n ]\n },\n \"priority\": {\n \"type\": \"string\",\n \"enum\": [\"p0\", \"p1\", \"p2\"]\n },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n }\n }\n}\n","frontmatter":{"$schema":"https://json-schema.org/draft/2020-12/schema","$id":"https://langbot.app/schemas/langbot-skills/suite.schema.json","title":"LangBot Skill Test Suite","type":"object","required":["id","title","description","type","priority","tags","cases"],"additionalProperties":true,"properties":{"id":{"type":"string","pattern":"^[a-z0-9][a-z0-9_-]*$"},"title":{"type":"string"},"description":{"type":"string"},"type":{"type":"string","enum":["smoke","regression","release_gate","exploratory","contract","performance","reliability","chaos","security"]},"priority":{"type":"string","enum":["p0","p1","p2"]},"tags":{"type":"array","items":{"type":"string"},"minItems":1},"cases":{"type":"array","items":{"type":"string"},"minItems":1}}},"isInternal":false,"tokens":325,"sizeBytes":1055},{"name":"troubleshooting.schema.json","path":"skills/schemas/troubleshooting.schema.json","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/schemas/troubleshooting.schema.json","title":"Schemas Skill","category":"anthropic-skill","format":"json","content":"{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/troubleshooting.schema.json\",\n \"title\": \"LangBot Skill Troubleshooting Entry\",\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"symptoms\", \"patterns\", \"likely_causes\", \"fix_steps\", \"verification\"],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"date\": {\n \"type\": \"string\"\n },\n \"category\": {\n \"type\": \"string\",\n \"enum\": [\"product\", \"env_issue\", \"external_dependency\", \"blocked\", \"flaky\"]\n },\n \"symptoms\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"likely_causes\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"fix_steps\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"verification\": {\n \"type\": \"string\"\n },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n}\n","frontmatter":{"$schema":"https://json-schema.org/draft/2020-12/schema","$id":"https://langbot.app/schemas/langbot-skills/troubleshooting.schema.json","title":"LangBot Skill Troubleshooting Entry","type":"object","required":["id","title","symptoms","patterns","likely_causes","fix_steps","verification"],"additionalProperties":true,"properties":{"id":{"type":"string","pattern":"^[a-z0-9][a-z0-9_-]*$"},"title":{"type":"string"},"date":{"type":"string"},"category":{"type":"string","enum":["product","env_issue","external_dependency","blocked","flaky"]},"symptoms":{"type":"array","items":{"type":"string"},"minItems":1},"patterns":{"type":"array","items":{"type":"string"},"minItems":1},"likely_causes":{"type":"array","items":{"type":"string"},"minItems":1},"fix_steps":{"type":"array","items":{"type":"string"},"minItems":1},"verification":{"type":"string"},"related_cases":{"type":"array","items":{"type":"string"}}}},"isInternal":false,"tokens":386,"sizeBytes":1243},{"name":"skills.index.json","path":"skills/skills.index.json","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills.index.json","title":"Skills Skill","category":"anthropic-skill","format":"json","content":"{\n \"generated_by\": \"lbs\",\n \"skills\": [\n {\n \"directory\": \"langbot-deploy\",\n \"name\": \"langbot-deploy\",\n \"description\": \"Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \\\"deploy langbot\\\", \\\"langbot docker\\\", \\\"langbot compose\\\", \\\"langbot kubernetes\\\", \\\"langbot config.yaml\\\", \\\"langbot box runtime\\\", \\\"langbot global api key\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-dev\",\n \"name\": \"langbot-dev\",\n \"description\": \"Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \\\"langbot backend\\\", \\\"langbot dev\\\", \\\"langbot api\\\", \\\"add langbot endpoint\\\", \\\"langbot migration\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-eba-adapter-dev\",\n \"name\": \"langbot-eba-adapter-dev\",\n \"description\": \"Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing.\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-env-setup\",\n \"name\": \"langbot-env-setup\",\n \"description\": \"Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths.\",\n \"references\": [\n \"references/browser-access-selection.md\",\n \"references/computer-use.md\",\n \"references/oauth-browser-profile.md\",\n \"references/playwright-mcp.md\",\n \"references/proxy.md\",\n \"references/service-startup.md\",\n \"references/wsl-notes.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-mcp-ops\",\n \"name\": \"langbot-mcp-ops\",\n \"description\": \"Operate a LangBot instance through its built-in MCP (Model Context Protocol) server. Use when an AI agent needs to manage LangBot — list/create/update/delete bots, agents, pipelines, models, knowledge bases, MCP servers, and skills — over MCP instead of raw HTTP. Covers the /mcp endpoint, API-key auth (web-UI lbk_ keys and the config.yaml global key), the tool surface, and client configuration. Triggers on \\\"langbot mcp\\\", \\\"manage langbot via mcp\\\", \\\"langbot /mcp\\\", \\\"langbot mcp server\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-plugin-dev\",\n \"name\": \"langbot-plugin-dev\",\n \"description\": \"Develop, debug, and test LangBot plugins. Use when creating new LangBot plugins, fixing plugin bugs, setting up a LangBot test environment, or testing plugins via WebSocket. Covers plugin component architecture (EventListener, Command, Tool), the plugin SDK API (invoke_llm, get_llm_models, send_message, plugin storage), common pitfalls, and automated WebSocket-based testing. Triggers on \\\"langbot plugin\\\", \\\"lbp\\\", \\\"GroupChatSummary\\\", \\\"plugin debug\\\", \\\"langbot test\\\".\",\n \"references\": [\n \"references/test-env-setup.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-skills-maintenance\",\n \"name\": \"langbot-skills-maintenance\",\n \"description\": \"Maintain the langbot-skills repository with low duplication. Use when adding, editing, or auditing LangBot skills, references, cases, troubleshooting entries, indexes, or periodic entropy-control checks for this skills repository.\",\n \"references\": [\n \"references/curation-workflow.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-space-ops\",\n \"name\": \"langbot-space-ops\",\n \"description\": \"Browse and search the LangBot Space marketplaces (plugins, MCP servers, skills) through the Space MCP server. Use when an AI agent needs to discover LangBot extensions on space.langbot.app over MCP. Covers the /mcp endpoint, Personal Access Token (PAT) auth, the tool surface, and client configuration. Triggers on \\\"langbot space mcp\\\", \\\"search langbot plugins\\\", \\\"langbot marketplace mcp\\\", \\\"space.langbot.app mcp\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-testing\",\n \"name\": \"langbot-testing\",\n \"description\": \"Test LangBot WebUI and core product flows with an automated browser and backend logs. Use when validating the configured LangBot frontend, pipeline Debug Chat, model provider setup and test buttons, bot and knowledge-base UI flows, or troubleshooting failed LangBot end-to-end tests.\",\n \"references\": [\n \"references/agent-runner-qa-workflow.md\",\n \"references/agent-runner-release-gate.md\",\n \"references/dify-agent-runner.md\",\n \"references/langrag-knowledge-base.md\",\n \"references/local-agent-runner-coverage.md\",\n \"references/local-agent-runner.md\",\n \"references/mcp-stdio-testing.md\",\n \"references/model-provider-testing.md\",\n \"references/performance-reliability-testing.md\",\n \"references/pipeline-debug-chat.md\",\n \"references/plugin-e2e-smoke.md\",\n \"references/sandbox-skill-authoring.md\",\n \"references/skill-all-tool-acceptance.md\",\n \"references/troubleshooting.md\",\n \"references/web-ui-testing.md\",\n \"references/workspace-release-testing.md\"\n ],\n \"cases\": [\n \"acp-agent-runner-debug-chat\",\n \"agent-run-ledger-audit\",\n \"agent-runner-async-db-readiness\",\n \"agent-runner-behavior-matrix\",\n \"agent-runner-fixture-contract\",\n \"agent-runner-health-visibility\",\n \"agent-runner-ledger-concurrency\",\n \"agent-runner-ledger-contention\",\n \"agent-runner-ledger-invariants\",\n \"agent-runner-ledger-stress\",\n \"agent-runner-live-install\",\n \"agent-runner-qa-debug-chat\",\n \"agent-runner-release-preflight\",\n \"agent-runner-runtime-chaos\",\n \"bot-event-routing-product-flow\",\n \"box-mcp-heartbeat-recovery\",\n \"claude-code-agent-debug-chat\",\n \"codex-agent-debug-chat\",\n \"dify-agent-debug-chat\",\n \"langbot-fake-provider-debug-chat-cross-pipeline-isolation\",\n \"langbot-fake-provider-debug-chat-fault-recovery\",\n \"langbot-fake-provider-debug-chat-load\",\n \"langbot-fake-provider-debug-chat-slow-load\",\n \"langbot-fault-taxonomy-contract\",\n \"langbot-live-backend-latency\",\n \"langbot-live-backend-log-health\",\n \"langbot-live-control-plane-api\",\n \"langbot-overhead-accounting-contract\",\n \"langbot-space-debug-chat-concurrency-smoke\",\n \"langrag-kb-retrieve\",\n \"langrag-parser-golden-e2e\",\n \"langrag-sentinel-kb-discover\",\n \"local-agent-basic-debug-chat\",\n \"local-agent-combo-rag-compaction-tool-debug-chat\",\n \"local-agent-complex-coding-task-debug-chat\",\n \"local-agent-context-compaction-debug-chat\",\n \"local-agent-effective-prompt-debug-chat\",\n \"local-agent-model-fallback-before-first-chunk-debug-chat\",\n \"local-agent-multimodal-debug-chat\",\n \"local-agent-multitool-rag-compaction-debug-chat\",\n \"local-agent-nonstreaming-debug-chat\",\n \"local-agent-parallel-tools-rag-compaction-debug-chat\",\n \"local-agent-plugin-tool-call-debug-chat\",\n \"local-agent-rag-debug-chat\",\n \"local-agent-rag-multimodal-debug-chat\",\n \"local-agent-steering-debug-chat\",\n \"local-agent-streaming-post-commit-failure-debug-chat\",\n \"local-agent-tool-error-recovery-debug-chat\",\n \"local-agent-tool-loop-limit-debug-chat\",\n \"mcp-stdio-register\",\n \"mcp-stdio-tool-call\",\n \"pipeline-debug-chat\",\n \"pipeline-debug-chat-performance\",\n \"plugin-e2e-smoke\",\n \"provider-deepseek\",\n \"qa-plugin-smoke-live-install\",\n \"sandbox-skill-authoring-e2e\",\n \"sandbox-skill-authoring-edit-existing-e2e\",\n \"skill-discovery-via-mcp-gateway\",\n \"webui-login-state\",\n \"wizard-onebot-agent-runtime\",\n \"wizard-runner-marketplace-catalog\",\n \"wizard-scenario-routing\",\n \"workspace-compatibility-preflight\",\n \"workspace-plugin-pages-smoke\",\n \"workspace-repository-contracts\"\n ],\n \"case_summaries\": [\n {\n \"id\": \"acp-agent-runner-debug-chat\",\n \"title\": \"ACP Runner can answer through Debug Chat using real remote Claude\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"acp\",\n \"claude\",\n \"external-runner\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_ACP_RUNNER_PIPELINE_URL\",\n \"LANGBOT_ACP_RUNNER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"agent-run-ledger-audit\",\n \"title\": \"Persisted Runner run ledger passes end-to-end invariants\",\n \"mode\": \"probe\",\n \"area\": \"agent\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"ledger\",\n \"audit\",\n \"tools\"\n ],\n \"automation\": \"scripts/e2e/agent-run-ledger-audit.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\",\n \"metrics\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-async-db-readiness\",\n \"title\": \"Runner async DB readiness probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"async-db\",\n \"aiosqlite\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-async-db-readiness.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-behavior-matrix\",\n \"title\": \"Runner deterministic behavior matrix probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"deterministic-runner\",\n \"protocol\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-behavior-matrix.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-fixture-contract\",\n \"title\": \"QA Runner fixture contract probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"fixture\",\n \"deterministic-runner\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-fixture-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-health-visibility\",\n \"title\": \"Agent configuration shows whether its selected runner is usable\",\n \"mode\": \"agent-browser\",\n \"area\": \"agent\",\n \"type\": \"feature\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"agent\",\n \"productization\",\n \"release-gate\"\n ],\n \"automation\": \"scripts/e2e/agent-runner-health-visibility.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-concurrency\",\n \"title\": \"Runner run ledger concurrency and auth pytest probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"pytest\",\n \"run-ledger\",\n \"concurrency\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-concurrency.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-contention\",\n \"title\": \"Runner ledger SQLite contention probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"stress\",\n \"ledger\",\n \"concurrency\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-contention.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-invariants\",\n \"title\": \"Runner ledger schema and status invariants probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"ledger\",\n \"invariant\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-invariants.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-stress\",\n \"title\": \"Runner ledger lightweight stress probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"stress\",\n \"ledger\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-stress.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-live-install\",\n \"title\": \"QA Runner package installs and registers in LangBot\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"plugin\",\n \"local-install\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/install-qa-plugin-smoke.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-qa-debug-chat\",\n \"title\": \"QA Runner returns deterministic output through Debug Chat\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"pipeline\",\n \"debug-chat\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"case:agent-runner-live-install\",\n \"node:scripts/e2e/ensure-qa-agent-runner-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_QA_RUNNER_PIPELINE_URL\",\n \"LANGBOT_QA_RUNNER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-release-preflight\",\n \"title\": \"Agent runner release gate preflight validates environment readiness\",\n \"mode\": \"agent-browser\",\n \"area\": \"release\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"release-gate\",\n \"preflight\",\n \"environment\"\n ],\n \"automation\": \"scripts/e2e/agent-runner-release-preflight.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-runtime-chaos\",\n \"title\": \"Runner SDK runtime chaos pytest probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"pytest\",\n \"runtime\",\n \"sdk\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-runtime-chaos.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"bot-event-routing-product-flow\",\n \"title\": \"Bot event routing and adapter input can be inspected from the WebUI\",\n \"mode\": \"agent-browser\",\n \"area\": \"bot\",\n \"type\": \"feature\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"bot\",\n \"eba\",\n \"event-routing\",\n \"productization\"\n ],\n \"automation\": \"scripts/e2e/bot-event-routing-product-flow.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"box-mcp-heartbeat-recovery\",\n \"title\": \"Box heartbeat restores an existing stdio MCP tool session\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"box\",\n \"mcp\",\n \"stdio\",\n \"reliability\",\n \"chaos\",\n \"fault-injection\"\n ],\n \"automation\": \"skills/langbot-testing/probes/box-mcp-heartbeat-recovery.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"metrics\",\n \"api_diagnostic\",\n \"resource_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"claude-code-agent-debug-chat\",\n \"title\": \"Claude Code Agent can retrieve LangBot history through MCP\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"claude-code\",\n \"external-runner\",\n \"mcp\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"codex-agent-debug-chat\",\n \"title\": \"Codex Agent can retrieve LangBot history through MCP\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"codex\",\n \"external-runner\",\n \"mcp\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"dify-agent-debug-chat\",\n \"title\": \"Dify Runner returns a response through Pipeline Debug Chat\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"provider\",\n \"priority\": \"p2\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"dify\",\n \"agent-runner\",\n \"pipeline\"\n ],\n \"automation\": \"\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-cross-pipeline-isolation\",\n \"title\": \"LangBot Debug Chat fake-provider cross-pipeline isolation probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"reliability\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"isolation\",\n \"concurrency\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-cross-pipeline-isolation.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-cross-pipelines.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_A_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_A_NAME\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_B_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_B_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-fault-recovery\",\n \"title\": \"LangBot Debug Chat fake-provider fault recovery probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"chaos\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"fault-injection\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-load\",\n \"title\": \"LangBot Debug Chat controlled fake-provider load probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"load\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-slow-load\",\n \"title\": \"LangBot Debug Chat slow fake-provider load probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"slow-provider\",\n \"load\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fault-taxonomy-contract\",\n \"title\": \"LangBot fault taxonomy and cleanup contract\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": true,\n \"tags\": [\n \"reliability\",\n \"chaos\",\n \"contract\",\n \"synthetic\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-fault-taxonomy-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-backend-latency\",\n \"title\": \"LangBot live backend basic latency probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"live-backend\",\n \"latency\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-backend-latency.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-backend-log-health\",\n \"title\": \"LangBot live backend log health probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"reliability\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"live-backend\",\n \"backend-log\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-backend-log-health.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"backend_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-control-plane-api\",\n \"title\": \"LangBot live control-plane API probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"reliability\",\n \"live-backend\",\n \"control-plane\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-control-plane-api.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-overhead-accounting-contract\",\n \"title\": \"LangBot overhead accounting metrics contract\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": true,\n \"tags\": [\n \"performance\",\n \"metrics\",\n \"contract\",\n \"synthetic\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-overhead-accounting-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"resource_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-space-debug-chat-concurrency-smoke\",\n \"title\": \"LangBot Debug Chat real Space-provider concurrency smoke\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"space\",\n \"live-provider\",\n \"smoke\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_PIPELINE_URL\",\n \"LANGBOT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_MODEL_UUID\",\n \"LANGBOT_E2E_MODEL_UUID\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langrag-kb-retrieve\",\n \"title\": \"LangRAG knowledge base ingests and retrieves a sentinel document\",\n \"mode\": \"agent-browser\",\n \"area\": \"knowledge\",\n \"type\": \"feature\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"langrag\",\n \"knowledge\",\n \"rag\"\n ],\n \"automation\": \"scripts/e2e/langrag-kb-retrieve.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"langrag-parser-golden-e2e\",\n \"title\": \"LangRAG and GeneralParsers retrieve a structured golden document\",\n \"mode\": \"agent-browser\",\n \"area\": \"knowledge\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"golden\",\n \"e2e\",\n \"langrag\",\n \"parser\",\n \"knowledge\"\n ],\n \"automation\": \"\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"langrag-sentinel-kb-discover\",\n \"title\": \"Existing LangRAG sentinel knowledge base is discoverable\",\n \"mode\": \"probe\",\n \"area\": \"knowledge\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"langrag\",\n \"knowledge\",\n \"rag\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/ensure-langrag-sentinel-kb.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-basic-debug-chat\",\n \"title\": \"Local Agent Debug Chat returns a deterministic streaming response\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"streaming\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-combo-rag-compaction-tool-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and plugin tool result in one Debug Chat run\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-complex-coding-task-debug-chat\",\n \"title\": \"Local Agent completes a multi-file coding task with iterative verification\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"agentic\",\n \"coding\",\n \"sandbox\",\n \"tools\",\n \"long-running\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/reset-complex-agent-task.mjs\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_COMPLEX_AGENT_WORKSPACE\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\",\n \"metrics\"\n ]\n },\n {\n \"id\": \"local-agent-context-compaction-debug-chat\",\n \"title\": \"Local Agent compacts long Debug Chat history and preserves older facts\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"context\",\n \"compaction\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-effective-prompt-debug-chat\",\n \"title\": \"Local Agent consumes host effective prompt after PromptPreProcessing\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"prompt\",\n \"plugin\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-model-fallback-before-first-chunk-debug-chat\",\n \"title\": \"Local Agent falls back when the primary model fails before streaming starts\",\n \"mode\": \"probe\",\n \"area\": \"pipeline\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"fallback\",\n \"streaming\",\n \"fake-provider\",\n \"fault-injection\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"local-agent-multimodal-debug-chat\",\n \"title\": \"Local Agent Debug Chat preserves uploaded image input\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"multimodal\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-multitool-rag-compaction-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and two-step plugin tool loop\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"tool-loop\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-nonstreaming-debug-chat\",\n \"title\": \"Local Agent Debug Chat returns a deterministic response with UI streaming disabled\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"non-streaming\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-parallel-tools-rag-compaction-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and parallel plugin tools\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"parallel-tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-plugin-tool-call-debug-chat\",\n \"title\": \"Local Agent can call a plugin-provided tool\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"plugin\",\n \"tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-rag-debug-chat\",\n \"title\": \"Local Agent Debug Chat answers from a LangRAG knowledge base\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-rag-multimodal-debug-chat\",\n \"title\": \"Local Agent preserves image input while using RAG context\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"rag\",\n \n\n","isInternal":false,"tokens":11130,"sizeBytes":50038},{"name":"SKILL.md","path":"skills/skills/langbot-deploy/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-deploy/SKILL.md","title":"langbot-deploy","category":"anthropic-skill","format":"markdown","content":"---\nname: langbot-deploy\ndescription: Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \"deploy langbot\", \"langbot docker\", \"langbot compose\", \"langbot kubernetes\", \"langbot config.yaml\", \"langbot box runtime\", \"langbot global api key\".\n---\n\n# LangBot Deployment & Configuration\n\nCovers running LangBot in production. For development see `langbot-dev`.\n\nRead https://langbot.app/llms.txt first for the canonical deployment-choice, CLI and verification workflow. The summary below applies when selecting hosting.\n\n## Deployment choice and optional managed hosting\n\nRespect the user's deployment preference and reuse an existing instance when available. Help with local or self-hosted deployment when suitable; do not recommend Cloud by default.\n\nOffer [LangBot Cloud](https://space.langbot.app/cloud) when the user wants managed hosting without maintaining a server, the local environment is unsuitable, or a deployment problem remains difficult after reasonable troubleshooting. Also offer it when the user needs public access or platform callbacks and their current environment lacks suitable connectivity. Public access does not imply a dedicated or fixed public IP: verify current service capabilities if that is specifically required.\n\nIf the user wants to support LangBot's continued development and open-source maintenance, mention the official hosted service as an optional way to do so, without pressure or repeated appeals. Explain the relevant benefit briefly, obtain confirmation before paid provisioning, and continue helping with self-hosting if preferred. Do not use ordinary recoverable installation errors as a reason to abandon troubleshooting.\n\n## Docker Compose (recommended)\n\n```bash\ngit clone https://github.com/langbot-app/LangBot\ncd LangBot/docker\n\n# Full stack (sandbox/Box + stdio MCP hosting + skill add/edit enabled)\ndocker compose --profile all up\n\n# Basic (no Box runtime)\ndocker compose up\n```\n\nThe `all` / `box` profile starts three services:\n\n- `langbot` — main app, serves API + UI on `:5300`.\n- `langbot_plugin_runtime` — plugin runtime (control `:5400`, debug `:5401`).\n- `langbot_box` — Box sandbox runtime (`:5410`). Uses the host Docker socket to\n spawn sandbox containers, so the **Box root host path and in-container path\n must be identical** (`BOX__LOCAL__HOST_ROOT=${LANGBOT_BOX_ROOT:-${PWD}/data/box}`).\n OSS allows its RPC and managed-process relay to run without a token when both\n sides leave `LANGBOT_BOX_CONTROL_TOKEN` unset. For an exposed endpoint, set\n the same value of at least 32 non-whitespace characters in both the LangBot\n and Box containers. Generate it once with `openssl rand -hex 32`; never put\n it in `box.runtime.endpoint` or commit it to config.\n\nA Compose deployment may optionally set\n`LANGBOT_PLUGIN_RUNTIME_CONTROL_TOKEN` on both `langbot` and\n`langbot_plugin_runtime` when port 5400 needs shared-secret protection. OSS\ndefaults to leaving it unset on both sides. If enabled, generate one value with\n`openssl rand -hex 32`; configuring only one side causes the control connection\nto fail. Kubernetes may use the `langbot-plugin-runtime-control` Secret shown in\n`docker/kubernetes.yaml`.\n\nWith Box off, the dashboard/skills list stays visible (read-only) but sandbox\ntools, skill add/edit, and stdio MCP are disabled. Set `box.enabled: false`\n(or `BOX__ENABLED=false`) to match.\n\n## Kubernetes\n\nSee `docker/kubernetes.yaml` and the deployment guide at\nhttps://langbot.app/docs. `docker/deploy-k8s-test.sh` is a test helper.\n\n## config.yaml (generated at `data/config.yaml` on first run)\n\nTop-level sections: `api`, `system`, `command`, `concurrency`, `proxy`,\n`database`, `vdb`, `storage`, `plugin`, `monitoring`, `box`, `space`.\n\nKey settings:\n\n| Key | Meaning |\n| --- | --- |\n| `api.port` | HTTP API + UI port (default 5300) |\n| `api.global_api_key` | **Global API key** for the HTTP API + MCP server. Non-empty = accepted with no login/DB record; no `lbk_` prefix required. Empty = disabled. Plaintext — trusted/internal only, serve over HTTPS. |\n| `plugin.runtime_ws_url` | Standalone plugin runtime WS URL (e.g. `ws://langbot_plugin_runtime:5400/control/ws`) |\n| `box.enabled` | Master switch for the Box sandbox runtime |\n| `box.backend` | `local` (Docker/nsjail autopick) / `docker` / `nsjail` / `e2b` / explicit unsafe `host`; env override `BOX__BACKEND` |\n| `box.runtime.endpoint` | External Box runtime URL (e.g. `ws://127.0.0.1:5410`); empty = local auto-managed |\n\nMany keys have `ENV__SUBKEY` overrides (e.g. `BOX__BACKEND`, `BOX__ENABLED`).\n\n## Runtimes & flags\n\n- LangBot started directly spawns the plugin runtime over **stdio**.\n- In containers it connects to a standalone runtime over **WebSocket**; start\n with `--standalone-runtime`.\n- Box has a parallel `--standalone-box` flag; the Docker box host is\n `langbot_box:5410`.\n- `box.backend: host` runs commands directly as the Box Runtime system user.\n It is never auto-selected, provides no sandbox isolation, and is only for\n trusted local development. A WebSocket-controlled host backend requires\n `LANGBOT_BOX_CONTROL_TOKEN`; local stdio control is allowed.\n\n## Global API key — enabling for agents/automation\n\n```yaml\n# data/config.yaml\napi:\n port: 5300\n global_api_key: 'a-strong-secret' # empty disables it\n```\n\nThis key authenticates both the HTTP API and the MCP server (`/mcp`) without a\nlogin session. See `langbot-mcp-ops` for using it, and `docs/API_KEY_AUTH.md`.\n\n## Pitfalls\n\n- \"No supported sandbox backend (Docker / nsjail / E2B)\" with Docker running\n usually means the user isn't in the `docker` group →\n `sudo usermod -aG docker ` and restart in a new shell.\n- Do not use `box.backend: host` as a production fallback. It cannot enforce\n image, filesystem, network, PID, CPU, memory, or storage isolation.\n- Box root host/container path mismatch breaks sandbox container creation.\n- Don't commit a non-empty `api.global_api_key` to version control.\n","frontmatter":{"name":"langbot-deploy","description":"Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \"deploy langbot\", \"langbot docker\", \"langbot compose\", \"langbot kubernetes\", \"langbot config.yaml\", \"langbot box runtime\", \"langbot global api key\"."},"isInternal":false,"tokens":1475,"sizeBytes":6189},{"name":"SKILL.md","path":"skills/skills/langbot-dev/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-dev/SKILL.md","title":"langbot-dev","category":"anthropic-skill","format":"markdown","content":"---\nname: langbot-dev\ndescription: Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \"langbot backend\", \"langbot dev\", \"langbot api\", \"add langbot endpoint\", \"langbot migration\".\n---\n\n# LangBot Core Development\n\nThis skill covers developing the LangBot core (the main repo), distinct from\nplugin development (see `langbot-plugin-dev`) and deployment (`langbot-deploy`).\n\n## Stack\n\n- **Backend**: Python `>=3.11,<4.0`, deps via `uv`. Framework: **Quart** (async\n Flask). Serves the HTTP API + pre-built web UI on `http://127.0.0.1:5300`.\n- **Frontend** (`web/`): **Vite + React Router 7 + shadcn/ui + Tailwind**,\n managed by `pnpm`. Dev server on `:3000`. (NOT Next.js — `dev` script is `vite`.)\n\n## Dev environment\n\n```bash\n# Backend\npip install uv\nuv sync --dev\nuv run main.py # API + UI on http://127.0.0.1:5300\n\n# Frontend (separate terminal)\ncd web\ncp .env.example .env\npnpm install\npnpm dev # http://127.0.0.1:3000 (reads VITE_API_BASE_URL)\n\n# Lint/format hooks (CI runs the same checks)\nuv run pre-commit install\n```\n\nFirst run generates `data/config.yaml`; DB defaults to SQLite (PostgreSQL\nsupported). Migrations run automatically on startup.\n\n## Repo layout (key paths)\n\n```\nsrc/langbot/\n├── __main__.py # entrypoint, CLI flags (--standalone-runtime/-box/--debug)\n├── pkg/\n│ ├── api/\n│ │ ├── http/ # Quart controllers + services\n│ │ │ ├── controller/groups/ # route groups (@group.group_class)\n│ │ │ └── service/ # business logic (called by controllers AND MCP)\n│ │ └── mcp/ # MCP server (server.py = tools, mount.py = ASGI dispatch)\n│ ├── core/ # app bootstrap, stages, task manager\n│ ├── platform/ provider/ pipeline/ plugin/ box/ skill/ rag/ vector/\n│ ├── command/ persistence/ storage/ config/ entity/ telemetry/\n│ └── templates/config.yaml # config template (top-level: api, system, plugin, box, space...)\n├── web/ # Vite SPA\n└── docker/ # compose deployment\n```\n\n## HTTP API auth model\n\nRoute auth is declared per-route via `AuthType` in\n`pkg/api/http/controller/group.py`:\n\n- `NONE` — public.\n- `USER_TOKEN` — web UI JWT (`Authorization: Bearer `).\n- `API_KEY` — `X-API-Key` or `Authorization: Bearer `.\n- `USER_TOKEN_OR_API_KEY` — either.\n\nAuthenticated routes receive an immutable `RequestContext` containing the\nprincipal, authorized Workspace membership, fixed-role permissions, instance,\nrequest id, and placement generation. A browser's `X-Workspace-Id` is only a\nselector and is always checked against the Account membership. Tenant services\nmust accept this context (or an explicit trusted execution context) and fail\nclosed when it is absent.\n\nAPI-key authentication accepts:\n\n1. the **global key** from `config.yaml` `api.global_api_key` only for a\n community instance with exactly one local Workspace, then\n2. **web-UI keys** whose one-time `lbk_` secret is stored only as a hash and is\n bound to one Workspace, explicit scopes, status, and optional expiry.\n\nAn API key derives its Workspace from the key record and ignores a caller's\nWorkspace selector. Public Bot/Webhook routes similarly derive Workspace from\nthe opaque owning resource rather than a header.\n\nRoute groups self-register via `@group.group_class(name, path)` and are\ndiscovered by `importutil.import_modules_in_pkg`.\n\n## Adding an API endpoint\n\n1. Add/extend a controller in `pkg/api/http/controller/groups/` and the matching\n service method in `pkg/api/http/service/`.\n2. Pick the right `AuthType`.\n3. **If the endpoint should be agent-accessible, add/adjust the matching MCP tool\n in `pkg/api/mcp/server.py` and update the `langbot-mcp-ops` skill.** API and\n MCP surface must stay aligned (see `AGENTS.md`).\n4. If `lbctl` uses the route or capability, check the separate\n [`langbot-cli`](https://github.com/langbot-app/langbot-cli) client for compatibility.\n5. Update `docs/service-api-openapi.json` if you maintain the OpenAPI overview.\n\n## Database migrations (Alembic)\n\nSingle migration set supports SQLite + PostgreSQL. Files in\n`src/langbot/pkg/persistence/alembic/versions/`.\n\n```bash\n# From project root (needs data/config.yaml)\nuv run python -m langbot.pkg.persistence.alembic_runner autogenerate \"description\"\n```\n\n## Standards\n\n- All code comments/docstrings in **English**; user-facing strings need **i18n**\n (`en_US` + `zh_Hans` minimum, `ja_JP` where present).\n- Consider toC and toB compatibility + security.\n- Commit format: `(): ` (feat/fix/docs/refactor/...).\n\n## Tests\n\n```bash\nuv run pytest tests/unit_tests -q # unit tests\nuv run pytest tests/unit_tests/api -q # API service tests\nuv run python tests/manual/mcp_smoke.py # MCP server e2e smoke\n```\n\n## See also\n\n- `langbot-plugin-dev` — plugin SDK / runtime development.\n- `langbot-testing` — WebUI/e2e QA harness (`bin/lbs`).\n- `langbot-deploy` — Docker/compose deployment + config.\n- `langbot-mcp-ops` — operating the LangBot MCP server.\n- [`langbot-cli`](https://github.com/langbot-app/langbot-cli) — `lbctl`, a standalone Service API client for managing running Workspaces.\n","frontmatter":{"name":"langbot-dev","description":"Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \"langbot backend\", \"langbot dev\", \"langbot api\", \"add langbot endpoint\", \"langbot migration\"."},"isInternal":false,"tokens":1440,"sizeBytes":5713},{"name":"SKILL.md","path":"skills/skills/langbot-eba-adapter-dev/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-eba-adapter-dev/SKILL.md","title":"langbot-eba-adapter-dev","category":"anthropic-skill","format":"markdown","content":"---\nname: langbot-eba-adapter-dev\ndescription: Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing.\n---\n\n# LangBot EBA Adapter Development\n\nUse this skill when implementing or reviewing a LangBot platform adapter under the Event-Based Agents architecture.\n\n## Controlling a running instance via MCP\n\nBeyond writing code, you can **drive a live LangBot instance over MCP** — no raw\nHTTP needed. Two MCP servers exist (both reuse existing API keys; see `AGENTS.md`):\n\n- **LangBot instance** — `http://:5300/mcp` (auth: web-UI `lbk_` key or the\n `api.global_api_key` from `config.yaml`). Manage bots, pipelines, models,\n knowledge bases, and skills. See the **`langbot-mcp-ops`** skill.\n- **LangBot Space marketplace** — `https://space.langbot.app/mcp` (auth: Personal\n Access Token). Search plugins / MCP servers / skills. See the\n **`langbot-space-ops`** skill.\n\n> Any change to an agent-accessible HTTP API endpoint must keep the matching MCP\n> tool and these skills in sync.\n\n## Core Rule\n\nDo not let platform-native event or message shapes leak into LangBot's common path. Each adapter must convert incoming SDK objects into unified EBA entities before dispatch:\n\n- Events: `langbot_plugin.api.entities.builtin.platform.events`\n- Message chains: `langbot_plugin.api.entities.builtin.platform.message.MessageChain`\n- Users/groups/members: `langbot_plugin.api.entities.builtin.platform.entities`\n- Raw platform objects may remain only in `source_platform_object` for debugging or platform-specific escape hatches.\n\n## Start Here\n\n1. Read the EBA design docs in `LangBot/docs/event-based-agents/`.\n2. Read the architecture-level acceptance checklist before writing or validating code:\n - `LangBot/docs/event-based-agents/adapters/acceptance-checklist.md`\n3. Read the current reference adapter before writing code. Prefer Telegram first:\n - `LangBot/src/langbot/pkg/platform/adapters/telegram/`\n - `LangBot/docs/event-based-agents/adapters/telegram.md`\n4. Read the legacy source adapter for the target platform:\n - `LangBot/src/langbot/pkg/platform/sources/.py`\n - `LangBot/src/langbot/pkg/platform/sources/.yaml`\n5. Inspect SDK entity definitions in `langbot-plugin-sdk/src/langbot_plugin/api/entities/builtin/platform/`.\n6. Search before assuming APIs. Platform SDKs change often.\n\n## Adapter Layout\n\nCreate one directory per adapter:\n\n```text\nLangBot/src/langbot/pkg/platform/adapters//\n├── __init__.py\n├── adapter.py\n├── api_impl.py\n├── event_converter.py\n├── manifest.yaml\n├── message_converter.py\n├── platform_api.py\n├── types.py\n└── .svg\n```\n\nAdd optional helpers such as `voice.py` only when the platform has a real domain-specific surface.\n\nEnsure `pyproject.toml` package data includes adapter assets:\n\n```toml\npackage-data = { \"langbot\" = [\"templates/**\", \"pkg/platform/sources/*\", \"pkg/platform/adapters/**\", ...] }\n```\n\n## Implementation Checklist\n\n- `manifest.yaml` declares `metadata.name`, config schema, supported events, common APIs, and platform-specific APIs.\n- `adapter.py` creates the platform client, subscribes to native events, filters self/bot loops where appropriate, calls `event_converter.target2yiri(...)`, then dispatches the EBA event.\n- `event_converter.py` maps native events to EBA event classes such as `MessageReceivedEvent`, `MessageEditedEvent`, `MessageDeletedEvent`, `MessageReactionEvent`, `MemberJoinedEvent`, `BotInvitedToGroupEvent`, and `PlatformSpecificEvent`.\n- `message_converter.py` maps native messages to `MessageChain`, and maps `MessageChain` back to the platform send format.\n- `api_impl.py` implements common EBA APIs: send, reply, edit, delete, forward, user/group/member lookup, moderation, upload/file URL, leave group.\n- `platform_api.py` keeps platform-specific calls behind `call_platform_api(action, params)`.\n- Unsupported common APIs must raise explicit SDK platform errors such as `NotSupportedError`; do not silently no-op.\n- Destructive APIs such as kick, ban, leave, delete, or moderation must be gated in live tests and documented.\n\n## Conversion Contract\n\nFor message events, the common shape should look like this regardless of platform:\n\n```python\nplatform_events.MessageReceivedEvent(\n type=\"message.received\",\n adapter_name=\"\",\n message_id=,\n message_chain=platform_message.MessageChain([...]),\n sender=platform_entities.User(...),\n chat_type=platform_entities.ChatType.PRIVATE or ChatType.GROUP,\n chat_id=,\n group=platform_entities.UserGroup(...) or None,\n source_platform_object=,\n)\n```\n\nMessage content should use common components:\n\n- `Source` for original message id/time when available.\n- `Plain` for text.\n- `At` / `AtAll` for mentions.\n- `Image`, `Voice`, `File` for media.\n- `Forward` only when the platform can represent or emulate it safely.\n\nIf a platform event cannot cleanly map to a common event, emit `PlatformSpecificEvent` with a compact `action` and structured `data`.\n\n## Unit Tests\n\nAdd focused tests under `LangBot/tests/unit_tests/platform/test__eba_adapter.py`.\n\nCover at least:\n\n- Manifest supported events match adapter `supported_events()`.\n- Manifest supported APIs match adapter `supported_apis()`.\n- Platform API map matches manifest actions.\n- Dispatcher chooses the most specific EBA listener.\n- Message converter maps every supported common component both directions where possible:\n - `Source`\n - `Plain`\n - `At`\n - `AtAll`\n - `Image`\n - `Voice`\n - `File`\n - `Quote`\n - `Face`\n - `Forward`\n - `Unknown`\n - mixed chains preserving order\n- Event converter maps message received/edited/deleted/reaction, raw uncached gateway events, member events, and bot join/leave events.\n- Send/reply methods pass correct platform kwargs and return `MessageResult`.\n\nRun the existing reference adapter tests too:\n\n```bash\ncd LangBot\nuv run pytest tests/unit_tests/platform/test__eba_adapter.py tests/unit_tests/platform/test_telegram_eba_adapter.py\nuv run python -m py_compile tests/e2e/live__eba_probe.py\ngit diff --check\n```\n\n## Live Test Workflow\n\nDirect adapter live probes are useful diagnostics, but they are not sufficient acceptance evidence for EBA. Treat `tests/e2e/live__eba_probe.py` as an auxiliary tool only. The final adapter record must distinguish:\n\n- `plugin-e2e-ui`: real SDK plugin through standalone runtime, LangBot core, adapter, and a real/simulator UI action. This can mark an inbound UI item complete.\n- `plugin-e2e-protocol`: real SDK plugin through standalone runtime, LangBot core, adapter, and a protocol-boundary injected event. This is useful evidence but must not be claimed as UI coverage.\n- `plugin-e2e-outbound`: real SDK plugin calls an API and the bot output is visible in the real/simulator UI. This can mark send/API coverage complete.\n- `adapter-live`: direct adapter probe connected to a real/simulator endpoint. This is auxiliary only.\n- `unit`: mocked conversion/API-shape coverage. This is auxiliary only.\n- `not-supported`: platform protocol or SDK has no equivalent. Must include the reason.\n- `blocked`: intended capability could not be verified. This is not complete.\n\nWrite a live probe in `LangBot/tests/e2e/live__eba_probe.py`. It should:\n\n1. Read token/client ids from environment variables or CLI args.\n2. Start the adapter directly.\n3. Register an EBA listener and write JSONL evidence to `LangBot/data/temp/`.\n4. Wait for a real user/platform event instead of fabricating the entrypoint.\n5. Exercise common APIs and `call_platform_api` actions.\n6. Observe returned gateway events for edit/delete/reaction/member/bot lifecycle where available.\n7. Print a summary containing passed, failed, skipped, and observed event types.\n8. Redact or avoid printing secrets.\n9. Keep destructive operations behind flags and run them last.\n\nUse Computer Use when the user asks for real platform end-to-end coverage. Actually send messages/click reactions in the platform UI or otherwise trigger real user-side events; do not replace that with unit tests.\n\nFor media/component acceptance, keep the direction and trigger source explicit:\n\n- Real inbound media only counts when a human-side platform UI or simulator UI sends the image/file/voice to the bot and the plugin JSONL records the corresponding common component.\n- Bot outbound media only proves `send_message`/adapter send conversion. It does not prove inbound conversion.\n- Protocol-boundary injection, such as sending a OneBot event directly into a reverse WebSocket adapter, is useful and should be labelled `plugin-e2e-protocol`, but it must not be reported as UI-level end-to-end media upload.\n- If the UI cannot send or upload the media, record the item as `blocked` with the exact client/simulator limitation.\n\n## Standalone Runtime + Plugin Test\n\nWhen validating the whole LangBot EBA path, test with the SDK standalone runtime and a real test plugin. This is the required acceptance path; direct adapter calls do not prove the EBA architecture path.\n\nThe required path is:\n\n```text\nReal platform / simulator UI\n -> platform SDK native event\n -> adapter event converter\n -> unified EBA event/entity/message types\n -> LangBot core event dispatch\n -> standalone SDK runtime\n -> real test plugin listener\n -> plugin calls platform APIs through SDK\n -> LangBot core API dispatch\n -> adapter API implementation\n -> real platform / simulator UI\n```\n\nTypical shape:\n\n```bash\n# Terminal 1, SDK repo\ncd langbot-plugin-sdk\nuv run python -m langbot_plugin.cli.__init__ rt \\\n --debug-only \\\n --ws-control-port 5400 \\\n --ws-debug-port 5401 \\\n --skip-deps-check\n\n# Terminal 2, LangBot repo\ncd LangBot\nexport PYTHONPATH=/absolute/path/to/langbot-plugin-sdk/src:${PYTHONPATH:-}\nuv run main.py --standalone-runtime\n\n# Terminal 3, plugin directory\nexport DEBUG_RUNTIME_WS_URL=ws://127.0.0.1:5401/plugin/ws\nexport EBA_PROBE_LOG=/absolute/path/to/LangBot/data/temp/_eba_plugin_probe.jsonl\nexport EBA_PROBE_API=1\nexport EBA_PROBE_COMPONENT_SWEEP=1\nexport EBA_PROBE_PLATFORM_API=1\nuv --project /absolute/path/to/langbot-plugin-sdk run python -m langbot_plugin.cli.__init__ run\n```\n\nUse an EBA probe plugin that subscribes to all relevant EBA event classes and runs SDK API calls after the first `MessageReceived`.\n\nThe plugin evidence should be JSONL and include:\n\n- event class and `event.type`\n- adapter name\n- chat type and chat ID\n- sender/user/group IDs with secrets redacted\n- `bot_uuid` and `adapter_name`, proving LangBot filled common routing fields before plugin dispatch\n- received `message_chain` component list\n- API action name, input summary, result or error\n- unsupported or blocked reason when an item is skipped\n\nFor full adapter acceptance, enable both probe sweeps:\n\n- `EBA_PROBE_COMPONENT_SWEEP=1` sends the required outbound message components through `send_message`.\n- `EBA_PROBE_PLATFORM_API=1` calls common safe APIs plus selected `call_platform_api` actions for the adapter.\n\nThe SDK must support `plugin.call_platform_api(bot_uuid, action, params)` for platform-specific acceptance. If the SDK cannot call a platform-specific action from the plugin, the adapter cannot be fully accepted even if direct adapter probes pass.\n\n## Required EBA Acceptance Coverage\n\nBefore marking an adapter migrated, fill out an adapter record against `LangBot/docs/event-based-agents/adapters/acceptance-checklist.md`.\n\nAt minimum, the record must cover these categories:\n\n- Message receive component tests through `plugin-e2e-ui`: `Source`, `Plain`, `At`, `AtAll`, `Image`, `Voice`, `File`, `Quote`, `Face`, `Forward`, `Unknown`, and mixed chains where the platform supports them. Protocol-only receive evidence must be labelled `plugin-e2e-protocol`.\n- Message send component tests through `plugin-e2e-outbound`: `Plain`, `At`, `AtAll`, `Image`, `Voice`, `File`, `Quote`, `Face`, `Forward`, and mixed chains where the platform supports them.\n- Every event declared in `manifest.yaml -> spec.supported_events`.\n- Every common API declared in `manifest.yaml -> spec.supported_apis.required` and `optional`.\n- Every action declared in `manifest.yaml -> spec.platform_specific_apis`.\n- Compatibility tests for manifest declarations, legacy message listener fallback, EBA listener specificity, bot self-message filtering, and `source_platform_object` reply/debug behavior.\n\nDo not declare an event or API in the manifest unless it has an implementation path and an acceptance entry. If a platform or simulator lacks a capability, document it as `not-supported` or `blocked` rather than silently omitting the test.\n\n## Common Pitfalls\n\n- `get_bots()` may return bot dictionaries, not UUID strings. Probe plugins should select an enabled dict and pass `bot[\"uuid\"]` to `get_bot_info()` and `send_message()`.\n- Make sure the probe subscribes to every event you claim to verify. Missing `MessageDeleted` subscription can make a working adapter look untested.\n- Some platforms emit both cached and raw gateway events, producing duplicate evidence for delete/reaction. Count this explicitly; do not treat duplicates as failure unless semantics differ.\n- Self-message filtering is platform-specific. Filter bot-originated `message.received` loops, but do not accidentally filter edit/delete events needed for bot-owned API probes.\n- Reaction events may be filtered for bot self reactions. To test user reaction add/remove, use real UI interaction or a real user token path if permitted.\n- File uploads usually happen as message attachments. A standalone `upload_file` API may need to be `NotSupportedError`.\n- Live probes should not leak bot tokens through command output, logs, docs, or final answers.\n- Discord requires privileged intents for message content and members. Missing intents can look like converter bugs.\n- Telegram Bot API exposes only limited member lists; document capability gaps.\n- Do not mark moderation APIs verified unless they ran against a disposable target member/bot.\n- If `leave_group` is tested, run it last because the test bot will be removed from the server/group.\n- Restore local LangBot DB/test state after live runs if you enabled temporary bots or changed plugin settings.\n\n## Documentation Record\n\nAdd or update `LangBot/docs/event-based-agents/adapters/.md` in the same style as Telegram:\n\n- Status and adapter directory.\n- Configuration table matching manifest fields.\n- Supported EBA event list.\n- Common API table with support and limitations.\n- `call_platform_api` action list.\n- Receive component table with evidence level per component.\n- Send component table with evidence level per component.\n- Event table with evidence level per event.\n- Common API table with evidence level per API.\n- Platform-specific API table with evidence level per action.\n- Live test record with exact date, endpoint/simulator, standalone runtime command, test plugin path/name, JSONL evidence path, channel/group type, observed events, APIs exercised, destructive operations, and skipped items.\n\nBe honest. Put untested or skipped APIs in the document with the reason. Do not imply full parity when a platform cannot provide the same information density.\n\n## Before Finishing\n\n- Run unit tests and compile the live probe.\n- Run the standalone runtime plugin E2E path for every required acceptance item that the platform supports.\n- Run `git diff --check`.\n- Summarize live JSONL evidence by event type.\n- Stop all long-running runtimes and probes.\n- Confirm no secrets are staged.\n- Leave unrelated untracked files alone.\n","frontmatter":{"name":"langbot-eba-adapter-dev","description":"Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing."},"isInternal":false,"tokens":3514,"sizeBytes":15943},{"name":"browser-access-selection.md","path":"skills/skills/langbot-env-setup/references/browser-access-selection.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-env-setup/references/browser-access-selection.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Browser Access Selection\n\nChoose the lightest browser-control path that can complete the task.\n\n## Decision Order\n\n1. If Codex Computer Use, Claude Computer Use, or another visible browser-control tool is available, use `computer-use.md`.\n2. If no computer-control tool is available but Playwright MCP is available, use `playwright-mcp.md`.\n3. If the browser session must survive restarts or OAuth login is required, also use `oauth-browser-profile.md`.\n4. If running under WSL, add `wsl-notes.md`.\n5. If external sites or model providers time out, add `proxy.md`.\n\n## Principle\n\nComputer Use and Playwright MCP are alternative browser-control paths. Both still need LangBot services to be reachable, so service checks stay in `service-startup.md`.\n","isInternal":false,"tokens":164,"sizeBytes":751},{"name":"computer-use.md","path":"skills/skills/langbot-env-setup/references/computer-use.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-env-setup/references/computer-use.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Computer Use Browser Path\n\nUse this path when Codex Computer Use, Claude Computer Use, or another agent-visible browser-control capability is available.\n\n## Why This Path Is Simpler\n\nComputer Use can interact with a visible browser directly, so it usually does not need Playwright MCP configuration or a separate MCP browser bridge.\n\n## Workflow\n\n1. Verify LangBot backend/frontend with `service-startup.md`.\n2. Open the WebUI in the controlled browser.\n3. If login is needed, let the user complete GitHub OAuth. Never handle credentials or 2FA.\n4. Keep the browser/profile available for later testing.\n5. Hand off to `langbot-testing` after the page shows the logged-in WebUI.\n\n## Still Required\n\n- Proxy may still be needed for GitHub OAuth or model provider tests. Use `proxy.md`.\n- Persisted profile details may still matter if the computer-control browser is restarted. Use `oauth-browser-profile.md` if login state must survive.\n","isInternal":false,"tokens":197,"sizeBytes":937},{"name":"oauth-browser-profile.md","path":"skills/skills/langbot-env-setup/references/oauth-browser-profile.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-env-setup/references/oauth-browser-profile.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# OAuth Browser Profile\n\nUse this reference when LangBot or LangBot Space needs GitHub OAuth login and the agent must reuse the authenticated browser state later.\n\nRead `skills/.env` first for `LANGBOT_BACKEND_URL`, `LANGBOT_FRONTEND_URL`, `LANGBOT_BROWSER_PROFILE`, `LANGBOT_CHROMIUM_EXECUTABLE`, and proxy defaults.\n\n## Rules\n\n- Never handle the user's GitHub password, passkey, recovery code, or 2FA secret.\n- Open a visible browser and let the user complete credential steps.\n- Reuse a fixed browser profile path.\n- Do not print token values. It is acceptable to report localStorage key names.\n\n## Manual Visible Login Flow\n\n1. Verify LangBot backend is reachable with `service-startup.md`.\n2. Launch a visible Chromium window with the persistent profile:\n\n```bash\nsetsid \"$LANGBOT_CHROMIUM_EXECUTABLE\" \\\n --no-sandbox \\\n --ozone-platform=x11 \\\n --user-data-dir=\"$LANGBOT_BROWSER_PROFILE\" \\\n --proxy-server=\"$LANGBOT_PROXY_SOCKS\" \\\n --proxy-bypass-list=\"$LANGBOT_NO_PROXY\" \\\n \"$LANGBOT_BACKEND_URL/login\" \\\n >/tmp/langbot-visible-chrome.log 2>&1 < /dev/null &\n```\n\n3. The user completes:\n\n```text\nLogin with Space -> Login with GitHub -> GitHub credentials / 2FA -> authorize\n```\n\n4. The agent can then reuse the same profile for automated checks.\n\n## Expected Successful State\n\nAfter login, LangBot should redirect away from `/login`, for example to a `/home/...` URL on the selected origin.\n\nExpected visible signals:\n\n```text\nLangBot\nDashboard\nHome\nBots\nPipelines\nKnowledge\nExtensions\n```\n\nExpected localStorage key names:\n\n```text\ntoken\nuserEmail\nlangbot_language\n```\n\nIf the user logged in on one origin but `LANGBOT_FRONTEND_URL` still shows `/login`, copy only the auth state needed between origins. Do not print token values.\n","isInternal":false,"tokens":420,"sizeBytes":1745},{"name":"playwright-mcp.md","path":"skills/skills/langbot-env-setup/references/playwright-mcp.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-env-setup/references/playwright-mcp.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Playwright MCP Browser Path\n\nUse this path when the agent needs browser automation but no Computer Use browser-control path is available.\n\n## Known Paths\n\n- Persistent browser profile: `LANGBOT_BROWSER_PROFILE` from `skills/.env.local`\n- Chromium executable: `LANGBOT_CHROMIUM_EXECUTABLE` from `skills/.env.local`\n- Codex MCP config: `$CODEX_HOME/config.toml` or the config path used by the active agent.\n\n## MCP Config\n\nKeep the profile path fixed so the agent can reuse authenticated state.\n\n```toml\n[mcp_servers.playwright]\ncommand = \"npx\"\nargs = [\"-y\", \"@playwright/mcp@latest\", \"--no-sandbox\", \"--executable-path\", \"\", \"--proxy-server\", \"\", \"--proxy-bypass\", \"localhost,127.0.0.1\", \"--user-data-dir\", \"\"]\n```\n\nAfter changing MCP config, restart Codex so the MCP server is relaunched with the new args.\n\n## Visible Login\n\nFor OAuth login, Playwright MCP's headless browser is not enough. Launch a visible browser with the same profile and let the user complete login. Use `oauth-browser-profile.md`.\n\n## Common Failures\n\n- MCP still uses old args after editing config: restart Codex or kill old `playwright-mcp` processes and restart the session.\n- Browser is headless during OAuth: use the visible login command from `oauth-browser-profile.md`.\n","isInternal":false,"tokens":318,"sizeBytes":1327},{"name":"proxy.md","path":"skills/skills/langbot-env-setup/references/proxy.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-env-setup/references/proxy.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Proxy Setup\n\nUse this reference when GitHub OAuth, package installation, model provider tests, or external API calls time out.\n\nRead defaults from `skills/.env` first.\n\n## Standard Local Proxy\n\n```bash\nexport HTTP_PROXY=\"$LANGBOT_PROXY_HTTP\"\nexport HTTPS_PROXY=\"$LANGBOT_PROXY_HTTP\"\nexport ALL_PROXY=\"$LANGBOT_PROXY_SOCKS\"\nexport http_proxy=\"$LANGBOT_PROXY_HTTP\"\nexport https_proxy=\"$LANGBOT_PROXY_HTTP\"\nexport all_proxy=\"$LANGBOT_PROXY_SOCKS\"\nexport NO_PROXY=\"$LANGBOT_NO_PROXY\"\nexport no_proxy=\"$LANGBOT_NO_PROXY\"\n```\n\n## Rule\n\nKeep uppercase and lowercase proxy variables consistent. Different libraries read different names.\n\n## Checks\n\n```bash\nenv | rg -i '^(http|https|all|no)_?proxy='\ncurl -I --max-time 8 --proxy \"$LANGBOT_PROXY_SOCKS\" https://github.com\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL\"\n```\n","isInternal":false,"tokens":202,"sizeBytes":813},{"name":"service-startup.md","path":"skills/skills/langbot-env-setup/references/service-startup.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-env-setup/references/service-startup.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Service Startup\n\nUse this reference for LangBot backend/frontend readiness checks regardless of OS or browser-control method. Read `skills/.env` first and override those defaults with user-provided values or detected running services.\n\n## Variables\n\n- `LANGBOT_REPO`\n- `LANGBOT_WEB_REPO`\n- `LANGBOT_BACKEND_URL`\n- `LANGBOT_FRONTEND_URL`\n- `LANGBOT_DEV_FRONTEND_URL`\n\n## Backend\n\nStart LangBot from the backend repo:\n\n```bash\ncd \"$LANGBOT_REPO\"\nuv run main.py\n```\n\nHealthy startup includes:\n\n```text\nRunning on http://0.0.0.0:\nConnected to plugin runtime.\nPlugin langbot-team/LocalAgent initialized\n```\n\nQuick check:\n\n```bash\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL/login\"\n```\n\nIf `bin/lbs env doctor` reports that `LANGBOT_BACKEND_URL` has no TCP listener,\nthe backend is not running at the configured host and port. A reachable\nstandalone frontend on `LANGBOT_FRONTEND_URL` does not prove backend readiness.\n\nPrefer a visible terminal session while debugging backend startup. Detached\nbackground startup methods can hide early process exits in local agent runs; if\nyou use one, immediately verify both the process and the listener:\n\n```bash\nps -eo pid,cmd | rg 'main.py|uv run main|langbot'\nss -ltnp | rg ':5300'\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL/login\"\n```\n\n## Frontend\n\nStart the new frontend from the web repo:\n\n```bash\ncd \"$LANGBOT_WEB_REPO\"\nVITE_API_BASE_URL=\"$LANGBOT_BACKEND_URL\" pnpm dev --host 0.0.0.0\n```\n\nHealthy startup includes:\n\n```text\nLocal: \n```\n\nQuick check:\n\n```bash\ncurl -I --max-time 3 \"$LANGBOT_FRONTEND_URL\"\n```\n\nIf `VITE_API_BASE_URL` is missing, Vite still serves the page but frontend API\ncalls may go to the frontend port instead of the backend port. That produces\nfalse browser failures in login, wizard, pipeline, and Debug Chat cases.\n\n## Completion Signal\n\nEnvironment setup is not complete until the required frontend/backend URLs are reachable and the chosen browser-control path can open the WebUI.\n","isInternal":false,"tokens":485,"sizeBytes":1976},{"name":"wsl-notes.md","path":"skills/skills/langbot-env-setup/references/wsl-notes.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-env-setup/references/wsl-notes.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# WSL Notes\n\nUse this reference only for WSL-specific details. Do not put generic LangBot startup or browser-login steps here.\n\n## Network\n\nGitHub login and model provider calls may require proxy access from WSL.\n\nWorking proxy form:\n\n```bash\nsocks5://127.0.0.1:7890\n```\n\nBypass local LangBot:\n\n```bash\nlocalhost,127.0.0.1\n```\n\nQuick checks:\n\n```bash\ncurl -I --max-time 8 --proxy socks5h://127.0.0.1:7890 https://github.com\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL\"\n```\n\n## Visible Browser\n\nIf OAuth requires a visible browser, WSL must have a usable display path. If a visible Chromium launch fails, check the local WSL GUI/X11 setup before changing LangBot config.\n\n## Common Failures\n\n- `ERR_NETWORK_CHANGED` or GitHub timeout: browser is not using the SOCKS proxy.\n- LangBot connection refused: backend is not running or not reachable from WSL.\n- User cannot type credentials: browser is headless or not visible to the user.\n","isInternal":false,"tokens":241,"sizeBytes":931},{"name":"SKILL.md","path":"skills/skills/langbot-env-setup/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-env-setup/SKILL.md","title":"langbot-env-setup","category":"anthropic-skill","format":"markdown","content":"---\nname: langbot-env-setup\ndescription: Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths.\n---\n\n# LangBot Environment Setup\n\nUse this skill when a task needs LangBot to be in a testable state before product testing or development verification.\n\n## Routing\n\n- **Shared local variables**: read `../.env` before using URL, path, browser profile, or proxy defaults.\n- **Always start here**: read `references/browser-access-selection.md` to choose the browser-control path.\n- **LangBot service checks and startup**: read `references/service-startup.md`.\n- **Computer Use available**: read `references/computer-use.md`. This path usually needs less browser/MCP setup.\n- **No Computer Use, browser automation required**: read `references/playwright-mcp.md`.\n- **GitHub OAuth or persisted login profile**: read `references/oauth-browser-profile.md`.\n- **WSL-specific notes**: read `references/wsl-notes.md` only when running under WSL.\n- **Proxy setup**: read `references/proxy.md` when external login, model provider tests, or package downloads time out.\n- **Headless-only automation**: use only after a profile already contains a valid LangBot login. Do not ask the agent to enter GitHub credentials or 2FA.\n\n## Rules\n\n- Never handle the user's GitHub password, passkey, recovery code, or 2FA secret.\n- For OAuth login, open a visible browser and let the user complete the credential steps.\n- Reuse a fixed browser profile path so the agent can later access the logged-in LangBot session.\n- Keep environment-specific paths and commands in `references/`, not in this file.\n- Treat environment setup as complete only after the target LangBot services are reachable and the browser profile can access the WebUI.\n","frontmatter":{"name":"langbot-env-setup","description":"Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths."},"isInternal":false,"tokens":413,"sizeBytes":1981},{"name":"SKILL.md","path":"skills/skills/langbot-mcp-ops/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-mcp-ops/SKILL.md","title":"langbot-mcp-ops","category":"anthropic-skill","format":"markdown","content":"---\nname: langbot-mcp-ops\ndescription: Operate a LangBot instance through its built-in MCP (Model Context Protocol) server. Use when an AI agent needs to manage LangBot — list/create/update/delete bots, agents, pipelines, models, knowledge bases, MCP servers, and skills — over MCP instead of raw HTTP. Covers the /mcp endpoint, API-key auth (web-UI lbk_ keys and the config.yaml global key), the tool surface, and client configuration. Triggers on \"langbot mcp\", \"manage langbot via mcp\", \"langbot /mcp\", \"langbot mcp server\".\n---\n\n# LangBot MCP Operations\n\nLangBot exposes an **MCP server** so AI agents can manage an instance\nprogrammatically. It mirrors a curated subset of the HTTP service API.\n\n## Endpoint\n\n```\nhttp://:5300/mcp\n```\n\nTransport: **streamable HTTP** (stateless, JSON responses). Same host/port as\nthe web UI and HTTP API.\n\n## Authentication\n\nReuses the same API keys as the HTTP API. Send either header:\n\n```\nX-API-Key: \n# or\nAuthorization: Bearer \n```\n\nTwo kinds of key are accepted:\n\n1. **Web-UI key** — created in the web UI (sidebar → API Keys), prefixed `lbk_`.\n The secret is shown once; only its SHA-256 hash is stored. Each key is bound\n to one Workspace and has explicit scopes, status, optional expiry, and\n last-used metadata. The key determines the Workspace; callers cannot switch\n it with `X-Workspace-Id`.\n2. **Global API key** — set in `data/config.yaml` under `api.global_api_key`.\n Requires no login session and no DB record; does not need the `lbk_` prefix.\n It is accepted only by a community instance with exactly one local\n Workspace and is disabled for SaaS multi-Workspace operation. Leave empty to\n disable. See the `langbot-deploy` skill for config details.\n\nInvalid, revoked, or expired keys get `401 Unauthorized`. A valid key whose\nscopes do not authorize a tool gets `403 Forbidden`.\n\nTo inspect key identity and permissions, call `GET /api/v1/system/context` with the API key.\n\n## Client configuration\n\n```json\n{\n \"mcpServers\": {\n \"langbot\": {\n \"url\": \"http://:5300/mcp\",\n \"headers\": { \"X-API-Key\": \"\" }\n }\n }\n}\n```\n\n## Tool surface\n\nSlack quick setup: create a disabled `slack-omni` bot draft first, then use\n`start_slack_setup` with the user's App Configuration access/refresh tokens.\nChoose `socket_mode=true` with a browser-reachable OAuth `redirect_url`, or\nprovide the public HTTPS `webhook_url` ending in `/bots/`.\nPoll `get_slack_setup_status` for the Slack installation authorization URL;\nthe user must authorize installation. On success, apply the returned `config`\nusing the normal bot update flow. Socket Mode also requires a separately\ngenerated `app_token` (`xapp-`, scope `connections:write`). Never log tokens.\n`cancel_slack_setup` removes the temporary session without deleting the Slack\napplication. Sessions expire after 15 minutes and are bound to the initiating\nWorkspace, principal, and placement generation. Restarting a failed setup can\ncreate another Slack application; inspect the returned `app_id` first.\n\nThe tools wrap the LangBot service layer. Current tools (v1):\n\n`create_pipeline` and HTTP `POST /api/v1/pipelines` preserve a supplied `config`\nand `extensions_preferences`; omit `config` to use installed Runner defaults.\nSelecting `ai.runner.id` never grants extension access: an empty `plugins` list\nwith `enable_all_plugins=false` still denies that plugin's Tool, EventListener\nand Command components, while allowing its selected Runner. Editing and copying\na pipeline retain these independent settings.\n\n| Tool | Purpose |\n| --- | --- |\n| `get_system_info` | Version, edition, instance id |\n| `list_bots` / `get_bot` / `create_bot` / `update_bot` / `delete_bot` | Manage messaging-platform bots (secrets redacted on read) |\n| `list_bot_event_route_statuses` | Inspect bot event-route runtime status |\n| `list_processors` / `get_processor` / `create_processor` / `update_processor` / `delete_processor` | Manage the peer Agent, Pipeline and Event processor types |\n| `get_processor_metadata` | Discover installed event-capable Runner components, schemas and supported event patterns. |\n| `list_processor_runs` / `get_processor_run_events` | Read one Agent or plugin processor run history and logs; paginate with `before_id` / `after_sequence`. |\n| `debug_agent` | Execute a synthetic Agent event (`processor_uuid`, `payload`); requires `runtime.operate`. Returns final text and up to 1000 execution events (thinking, text, tool arguments/results). Platform tools use Mock; other configured tools execute normally. Optional `payload.mock`: `errors`/`results` keyed by platform tool name, `unsupported_apis` lists unavailable platform APIs. |\n| `list_pipelines` / `get_pipeline` / `create_pipeline` / `update_pipeline` / `delete_pipeline` | Manage pipelines |\n| `list_llm_models` / `get_llm_model` / `list_embedding_models` / `list_model_providers` | Inspect models & providers |\n| `list_knowledge_bases` / `get_knowledge_base` / `retrieve_knowledge_base` | RAG knowledge bases (incl. semantic search) |\n| `list_mcp_servers` | External MCP servers LangBot connects to (as a client) |\n| `list_skills` / `get_skill` | Installed skills |\n| `list_knowledge_engines` / `get_knowledge_engine_schema` / `list_knowledge_parsers` | Discover RAG configuration |\n| `get_pipeline_extensions` / `update_pipeline_extensions` | Read or completely replace extension bindings; all lists and switches required |\n| `run_pipeline` | One fresh-session turn; requires `runtime.operate`, executes configured models/tools, never auto-retry an unknown outcome |\n| `get_monitoring_records` / `get_monitoring_details` | Bounded Workspace records and existing message/session details |\n| `get_sandbox_diagnostics` | Read status (`resource.view`), sessions/errors (`audit.view`); managed sandbox admission still applies |\n\nMutating tools (`create_*`, `update_*`) take a JSON object matching the same\nshape as the corresponding HTTP API request body. Discover resources with the\n`list_*` / `get_*` tools before mutating; identifiers are UUIDs. Reads require\n`resource.view`; mutations require `resource.manage`. All service calls inherit\nthe immutable Workspace context authenticated at the MCP transport boundary.\nPass `is_default: true` to `create_pipeline` only when the Workspace does not\nalready have a default pipeline.\n\n## How to use\n\nPipeline plugin extension switches authorize only EventListener, Tool and Command\ncomponents. Select the Runner independently in AI settings. An empty extension\nallowlist does not prevent saving, copying or running that Runner. Never enable\nall plugins or add its plugin to the allowlist as a Runner workaround: that also\nauthorizes its other components. The Runner plugin must still be installed and\nenabled in the authenticated Workspace, with an available Plugin Runtime.\n\n1. Get an API key (web UI key, or set `api.global_api_key` in config.yaml).\n2. Point your MCP client at `http://:5300/mcp` with the key header.\n3. Call `get_system_info` to confirm connectivity.\n4. Use `list_*` tools to discover, then `get_*` / `create_*` / `update_*` /\n `delete_*` as needed.\n\n## ChatGPT / Codex subscription providers\n\n`list_model_providers` can return the `openai-codex` requester. Its OAuth\ncredentials are server-only and are not provider API keys. Never ask a user\nto paste ChatGPT access tokens, refresh tokens, or a Codex auth cache into an\nMCP tool or model configuration.\n\nA human connects or disconnects the subscription through **Models → provider\nsettings** in the LangBot web UI. The provider-scoped `/codex/*` authentication\nroutes deliberately require a browser-user session and are not exposed as MCP\ntools or authorized by a LangBot API key. Once connected, models are managed\nand selected through the normal provider/model workflow. A disconnected\nprovider must be reauthorized; do not silently replace it with API-key billing.\n\nSee [ChatGPT / Codex subscription](../../../docs/CODEX_SUBSCRIPTION.md) for setup,\nusage limits, and the personal-account versus shared-service boundary.\n\n## Provider deletion\n\nThe curated MCP surface currently lists providers but has no provider-deletion\ntool. In the web UI, **Edit Provider → Delete** asks for confirmation before\nremoving that provider and all its LLM, embedding, and rerank models. This is\nirreversible; never interpret a request to edit a provider as authorization to\ndelete it.\n\nThe equivalent HTTP operation is\n`DELETE /api/v1/provider/providers/{uuid}?cascade=true`, requiring\n`resource.manage` in the authenticated Workspace. Omitting `cascade` preserves\nthe existing refusal to delete providers that still have models. Cloud-managed\nproviders remain protected. Cascade deletion removes stored Codex authorization\nstate as well; it is not the same operation as disconnecting an account.\n\n## Implementation & maintenance (for LangBot developers)\n\n- Server: `src/langbot/pkg/api/mcp/server.py` (FastMCP). Tools call the service\n layer directly, so the MCP surface stays aligned with the API.\n- Mount: `src/langbot/pkg/api/mcp/mount.py` — an ASGI dispatcher fronting Quart,\n authenticating `/mcp` requests, running the streamable-HTTP session manager.\n- Smoke test: `tests/manual/mcp_smoke.py`.\n\n> When you add, remove, or change an HTTP API endpoint that should be\n> agent-accessible, update the corresponding MCP tool **and** this skill. The\n> MCP tool surface and the API must stay aligned (see `AGENTS.md`).\n\n## Pitfalls\n\n- `/mcp` is the **server** LangBot exposes. The `/api/v1/mcp` routes are the\n **client** side (managing external MCP servers LangBot connects to). Don't\n confuse them.\n- A `401` means the key is wrong, missing, revoked, expired, or (for the global\n key) `api.global_api_key` is empty or the instance is not an OSS singleton.\n- A `403` means the key is valid but lacks the permission required by the tool.\n- The global key is plaintext in config.yaml — only enable it on trusted/internal\n deployments and serve over HTTPS.\n\n## Default message routes\n\n`create_bot` without non-empty `event_bindings` adds two explicit message routes:\nprivate chat (`chat_type eq person`) and group chat (`chat_type eq group` plus\n`is_at_bot eq true`). Both use the Workspace's default Pipeline, or the first\navailable Pipeline. If none exists, first-bot setup creates one with the full\ndefault config. Creating a Pipeline via `create_agent(kind=\"pipeline\")` also uses\nthat full default config. Configure an authorized model before testing replies.\n\nExplicit routes on create are preserved; `update_bot` never automatically\nreplaces manual routes, including an intentionally empty list. The WebUI message\ntemplate uses the same private/group conditions and preserves existing routes\nwhen applied repeatedly. Unmentioned group messages do not enter the Pipeline.\nAn existing broad route can still match first; inspect route order and diagnostics\nbefore changing a member's configuration.\n\n## Event processors\n\nCreate a processor with `kind: \"event_processor\"` and basic information. Without\na component it supports no events. Discover installed components with\n`get_processor_metadata`, then use `update_processor` with `component_ref` and\noptional `parameters`. API callers may also supply these when creating an instance.\nBind an instance by updating the bot's `plugin_processors` array with\n`{\"processor_uuid\": \"\", \"enabled\": true}`. This replaces the full\nsubscription list; preserve bindings you want to keep. Do not add plugin processors\nto `event_bindings`, which remains exclusive Agent/Pipeline routing.\nEach enabled subscription independently receives the installed Runner's declared\nevents. Slow or failed subscribers do not prevent other subscribers or the primary\nroute from executing. Installation alone never activates a handler. Reusing an\ninstance shares its configuration and runtime state. Use a separate instance for\nindependent settings. Optional plugin behavior belongs in the Runner config schema.\n`debug_agent` accepts the complete typed event in `payload.data` for this kind.\nLegacy EventListener plugins remain in the Pipeline lifecycle.\n\n`list_processor_runs` includes `created_at_ms`, `started_at_ms`, and\n`finished_at_ms`: Host lifecycle times in epoch milliseconds. Use the start and finish times for elapsed processing time; select a run and call `get_processor_run_events` for its\nlogs and action results. These times are not internal plugin profiling data.\n\n## Unified execution monitoring\n\nUse `get_inflight_executions` for a bounded current-workspace snapshot of active\nand recently started executions, including `progress_event`. Progress is a\nreported stage, not an estimated completion percentage. The UI uses the shared\n`GET /api/v1/monitoring/in-flight/stream` SSE feed instead of polling per viewer.\n\nUse `get_monitoring_executions` for the execution list and\nits legacy `pipeline_ids` parameter to filter any processor kind (Agent,\nPipeline or event processor); the summary uses the same processor scope. Use\n`get_monitoring_execution_detail` with `source=auto` to resolve a run,\nmessage or event identifier outside the current list page. The detail exposes\n`inputs`, `outputs` (generated content), `deliveries` (recorded platform sends),\n`events`, `llm_calls`, `tool_calls`, `errors`, `related`, and `conversation` in\n`pages`. Follow each section's `has_more` and `next_offset` independently.\nStart with `section=overview` to load only inputs, outputs, deliveries and errors.\nRequest `conversation`, `related`, calls and events separately when needed;\nomitting `section` retains the complete trace response for existing consumers.\nConversation history supplies context; historical messages without explicit\nlinks must not be asserted to belong to the selected execution. Events without\na processor run use `source=event`. All lookups remain Workspace-scoped.\n\nMonitoring record filters, except for `kind=\"sessions\"`, accept `mode` (`all`, `real`, `debug`) and\n`execution_statuses` (normalized execution statuses). These select the owning\nexecution, not the individual model/tool call outcome. Calls without a recorded\nexecution link are excluded when an execution filter is active.\n\nSession lists support `bot_ids`, `pipeline_ids`, `start_time`, `end_time`,\n`user_query`, `is_active`, and pagination. For execution status or real/debug\nfilters, use `get_monitoring_executions`; HTTP `/monitoring/sessions` rejects\nthem. In HTTP `/monitoring/data`, `sessions` and `totalCount.sessions` use only\nbot/pipeline/time filters; execution filters apply to the other records.\n\n### Workspace message language\n\n`get_workspace_message_language` reads the current Workspace's backend language\n(`workspace.view`). `set_workspace_message_language(language)` requires\n`workspace.update`; supported values are `en_US`, `zh_Hans`, `zh_Hant`, `ja_JP`,\n`es_ES`, `ru_RU`, `th_TH`, `vi_VN`. HTTP equivalents are GET/PUT\n`/api/v1/workspaces/current/language` (PUT body: `{\"language\":\"ja_JP\"}`).\nThe API key's Workspace remains authoritative. The setting controls new backend\nmessages and default bot notices, independently of personal UI language.\nPipeline `output.misc` hint fields override defaults; empty values inherit this\nsetting. Existing custom text and historical logs are not rewritten.\n\n### Workspace default LLM model\n\n`get_starred_model` reads the Workspace-wide favorite. `set_starred_model`\nrequires `provider_secret.manage` and replaces the single starred model; pass\n`null` to clear it. The model must belong to the current Workspace.\n`get_default_model` resolves the favorite first, then the Space wizard chat\nrecommendation only when LangBot Models is enabled and the Workspace owner is\nbound to a LangBot Account. It returns a nullable `uuid`.\nNew processor LLM selector defaults use this preference; existing configurations\nare not rewritten. Embedding and rerank models are not eligible.\n\n### Reset a bot session context\n\nUse `reset_session_context(bot_id, session_id)` only when a user asks to start a session afresh. It requires `resource.manage`, preserves monitoring records, and refuses sessions with active tasks. It clears conversation runner state and excludes earlier transcript entries from subsequent model context. Files and long-term memory are not removed.\n\n## Cloud plan platform restrictions\n\nBot creation and enabling use the same Workspace entitlement policy as HTTP. A disabled platform is rejected even through MCP or plugin calls; do not retry using an Omni or legacy alias to bypass the restriction. The Workspace-scoped HTTP adapter catalog omits disabled platforms.\n\n## LLM request diagnostics\n\n`get_monitoring_records(kind=\"llm_calls\")` includes nullable `details` on each call.\nRead `details.timing`, `details.usage`, and `details.error` to distinguish provider\nlatency, retry waits, cached/reasoning tokens, and failures. Timing values are\nmilliseconds since invocation start; `first_event_ms` includes reasoning/control\nframes, while `first_content_ms` measures the first upstream content delta.\n`first_visible_chunk_ms` may contain displayed reasoning, not just answer text.\nOld records have no details. Do not equate total duration with provider TTFT.\nSee `docs/LLM_REQUEST_POLICY.md` for policy precedence and gateway header behavior.\n\n`get_llm_model` also exposes optional `reasoning_capabilities.budget_tokens`\nmin/max bounds. Numeric thinking budgets use `reasoning_config.budget_tokens`\nwith level `enabled`; do not combine budgets with effort levels or native thinking\narguments. Existing Runner reasoning overrides take precedence. Prompt caching\nis opt-in through model/call `extra_args.prompt_caching`; use provider-specific\nmodes documented in `docs/LLM_REQUEST_POLICY.md`. Cache keys are routing hints,\nnot evidence of hits; read the actual usage counters. `preparation_ms` measures\nrequest construction and `preparation_cache_hits` counts retry preparation reuse.\n`transport_setup_ms` measures managed pool acquisition, including first-use TLS\ninitialization; it is not network TTFT. See `docs/LLM_LATENCY_BENCHMARK.md` for\ncontrolled request-layer measurements and their scope.\n","frontmatter":{"name":"langbot-mcp-ops","description":"Operate a LangBot instance through its built-in MCP (Model Context Protocol) server. Use when an AI agent needs to manage LangBot — list/create/update/delete bots, agents, pipelines, models, knowledge bases, MCP servers, and skills — over MCP instead of raw HTTP. Covers the /mcp endpoint, API-key auth (web-UI lbk_ keys and the config.yaml global key), the tool surface, and client configuration. Triggers on \"langbot mcp\", \"manage langbot via mcp\", \"langbot /mcp\", \"langbot mcp server\"."},"isInternal":false,"tokens":4012,"sizeBytes":18176},{"name":"test-env-setup.md","path":"skills/skills/langbot-plugin-dev/references/test-env-setup.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-plugin-dev/references/test-env-setup.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Test Environment Setup\n\n## Docker Compose (GitOps)\n\nCreate in `server-deploy` repo under `servers//langbot-test/docker-compose.yaml`:\n\n```yaml\nversion: \"3\"\nservices:\n langbot_plugin_runtime:\n image: rockchin/langbot:latest\n container_name: langbot-test-runtime\n volumes:\n - /opt/docker-data/langbot-test/data/plugins:/app/data/plugins\n ports:\n - \"5411:5401\"\n restart: on-failure\n environment:\n - TZ=Asia/Shanghai\n command: [\"uv\", \"run\", \"--no-sync\", \"-m\", \"langbot_plugin.cli.__init__\", \"rt\"]\n networks:\n - langbot_test_network\n\n langbot:\n image: rockchin/langbot:latest\n container_name: langbot-test\n volumes:\n - /opt/docker-data/langbot-test/data:/app/data\n ports:\n - \"5310:5300\"\n restart: on-failure\n depends_on:\n - langbot_plugin_runtime\n environment:\n - TZ=Asia/Shanghai\n networks:\n - langbot_test_network\n\nnetworks:\n langbot_test_network:\n driver: bridge\n```\n\n## Post-Deploy Configuration\n\nAfter first start, LangBot auto-generates `data/config.yaml`. You need to update `plugin.runtime_ws_url` to match the runtime container name:\n\n```bash\n# On the host, edit config\nsed -i 's|ws://localhost:5400/control/ws|ws://langbot-test-runtime:5400/control/ws|' \\\n /opt/docker-data/langbot-test/data/config.yaml\ndocker restart langbot-test\n```\n\n## Installing a Plugin\n\nCopy plugin directory to `data/plugins/` on the host:\n\n```bash\nscp -r MyPlugin/ user@host:/opt/docker-data/langbot-test/data/plugins/MyPlugin/\ndocker restart langbot-test-runtime # Runtime picks up new plugins on restart\n```\n\n## Caddy Reverse Proxy (Optional)\n\nIf testing externally, add to Caddyfile on the same host:\n\n```\nlangbot-test.example.com {\n reverse_proxy langbot-test:5300\n}\n```\n\nThen reload: `docker exec caddy caddy reload --config /etc/caddy/Caddyfile`\n\nThe WebSocket endpoint works through Caddy without special config.\n\n## WebSocket Test Script (Node.js)\n\n```javascript\nconst WebSocket = require('ws');\n\nconst PIPELINE_UUID = '';\nconst BASE = 'wss://langbot-test.example.com';\nconst URL = `${BASE}/api/v1/pipelines/${PIPELINE_UUID}/ws/connect?session_type=group`;\n\nconst ws = new WebSocket(URL, {\n headers: { Origin: BASE }\n});\n\nconst send = (text) => {\n ws.send(JSON.stringify({\n type: 'message',\n message: [{ type: 'Plain', text }]\n }));\n console.log('[SENT]', text);\n};\n\nws.on('message', (data) => {\n const msg = JSON.parse(data.toString());\n if (msg.type === 'connected') {\n console.log('Connected!');\n // Send test messages\n send('Message 1');\n setTimeout(() => send('Message 2'), 500);\n setTimeout(() => send('!summary'), 2000);\n } else if (msg.type === 'response' && msg.data?.is_final) {\n console.log('[BOT]', msg.data.content);\n }\n});\n\nws.on('error', (e) => console.error('Error:', e.message));\nsetTimeout(() => { ws.close(); process.exit(); }, 60000);\n```\n\nRequires: `npm install ws`\n","isInternal":false,"tokens":770,"sizeBytes":2944},{"name":"SKILL.md","path":"skills/skills/langbot-plugin-dev/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-plugin-dev/SKILL.md","title":"langbot-plugin-dev","category":"anthropic-skill","format":"markdown","content":"---\nname: langbot-plugin-dev\ndescription: Develop, debug, and test LangBot plugins. Use when creating new LangBot plugins, fixing plugin bugs, setting up a LangBot test environment, or testing plugins via WebSocket. Covers plugin component architecture (EventListener, Command, Tool), the plugin SDK API (invoke_llm, get_llm_models, send_message, plugin storage), common pitfalls, and automated WebSocket-based testing. Triggers on \"langbot plugin\", \"lbp\", \"GroupChatSummary\", \"plugin debug\", \"langbot test\".\n---\n\n# LangBot Plugin Development & Debugging\n\n## Controlling a running instance via MCP\n\nBeyond writing code, you can **drive a live LangBot instance over MCP** — no raw\nHTTP needed. Two MCP servers exist (both reuse existing API keys; see `AGENTS.md`):\n\n- **LangBot instance** — `http://:5300/mcp` (auth: web-UI `lbk_` key or the\n `api.global_api_key` from `config.yaml`). Manage bots, pipelines, models,\n knowledge bases, and skills. See the **`langbot-mcp-ops`** skill.\n- **LangBot Space marketplace** — `https://space.langbot.app/mcp` (auth: Personal\n Access Token). Search plugins / MCP servers / skills. See the\n **`langbot-space-ops`** skill.\n\n> Any change to an agent-accessible HTTP API endpoint must keep the matching MCP\n> tool and these skills in sync.\n\n## Plugin Architecture\n\nA LangBot plugin consists of:\n\n```\nMyPlugin/\n├── manifest.yaml # Plugin metadata, config schema\n├── main.py # BasePlugin subclass (entry point, shared state)\n├── components/\n│ ├── event_listener/ # Hook pipeline events\n│ │ ├── collector.yaml\n│ │ └── collector.py\n│ ├── commands/ # !command handlers\n│ │ ├── mycommand.yaml\n│ │ └── mycommand.py\n│ └── tools/ # LLM function-call tools\n│ ├── mytool.yaml\n│ └── mytool.py\n```\n\nEach component has a `.yaml` (metadata) and `.py` (implementation).\n\n## README & i18n convention (enforced on the marketplace)\n\nA plugin published to LangBot Space serves a localized README on its detail page.\nThe resolver (`langbot-space` `PluginService.GetPluginREADME`) works like this:\n\n- **Root `README.md` MUST be in English.** It is the default and the fallback —\n when no per-language README matches the viewer's locale, the page serves the\n root `README.md`. A non-English root README makes the English/default view show\n the wrong language.\n- **All other languages live under `readme/README_{lang}.md`** — e.g.\n `readme/README_zh_Hans.md`, `readme/README_ja_JP.md`. The 8 supported locales:\n `en_US, zh_Hans, zh_Hant, ja_JP, th_TH, vi_VN, es_ES, ru_RU`.\n- `manifest.yaml` `metadata.label` / `metadata.description` should carry the same\n 8-locale i18n set (`repository` must be a real, alive URL).\n\n```\nMyPlugin/\n├── manifest.yaml\n├── README.md # English (default + fallback) — REQUIRED, must be English\n└── readme/\n ├── README_zh_Hans.md\n ├── README_zh_Hant.md\n ├── README_ja_JP.md\n ├── README_th_TH.md\n ├── README_vi_VN.md\n ├── README_es_ES.md\n └── README_ru_RU.md\n```\n\n`manifest.yaml` (incl. `repository`) is the source of truth — the marketplace\nsyncs from it, so edit the package and re-publish rather than patching live data.\n\n## Critical SDK Pitfalls\n\n### 1. MessageChain is a RootModel — iterate directly\n\n```python\n# ❌ WRONG — MessageChain has no .components attribute\nfor component in event.message_chain.components:\n\n# ✅ CORRECT — MessageChain is a Pydantic RootModel, iterate directly\nfor component in event.message_chain:\n```\n\n### 2. Message.content must be `list[ContentElement]` or `str`, not a single ContentElement\n\n```python\nfrom langbot_plugin.api.entities.builtin.provider import message as provider_message\n\n# ❌ WRONG — single ContentElement\nMessage(role=\"user\", content=ContentElement.from_text(\"hello\"))\n\n# ✅ CORRECT — list of ContentElement\nMessage(role=\"user\", content=[ContentElement.from_text(\"hello\")])\n\n# ✅ ALSO CORRECT — plain string\nMessage(role=\"user\", content=\"hello\")\n```\n\n### 3. invoke_llm does NOT accept timeout\n\n```python\n# ❌ WRONG\nawait self.invoke_llm(llm_model_uuid=uuid, messages=msgs, timeout=60)\n\n# ✅ CORRECT\nawait self.invoke_llm(llm_model_uuid=uuid, messages=msgs)\n```\n\n### 4. invoke_llm response.content can be str OR list\n\n```python\nresponse = await self.invoke_llm(...)\nif response.content:\n if isinstance(response.content, str):\n return response.content\n elif isinstance(response.content, list):\n parts = [e.text for e in response.content if hasattr(e, \"text\") and e.text]\n return \"\\n\".join(parts)\n```\n\n### 5. get_llm_models() returns UUIDs\n\n```python\n# Returns list[str] of model UUIDs\nmodels = await self.get_llm_models()\nmodel_uuid = models[0] # First available model UUID\n```\n\n**Known bug (v4.9.3):** The host handler may return `list[dict]` instead of `list[str]`. If you hit `TypeError: unhashable type: 'dict'` in `invoke_llm`, the fix is in `LangBot/src/langbot/pkg/plugin/handler.py` — change `'llm_models': llm_models` to `'llm_models': [m['uuid'] for m in llm_models]`.\n\n### 6. invoke_llm parameter is `llm_model_uuid`, NOT `model_uuid`\n\n```python\n# ❌ WRONG — will throw \"got an unexpected keyword argument\"\nawait self.invoke_llm(messages=msgs, model_uuid=uuid)\n\n# ✅ CORRECT\nawait self.invoke_llm(messages=msgs, llm_model_uuid=uuid)\n```\n\n### 7. prevent_default() alone does NOT block LLM response\n\nTo fully prevent the default LLM pipeline from responding when your EventListener handles the message, you must call **both**:\n\n```python\nevent_context.prevent_default() # Block default behavior\nevent_context.prevent_postorder() # Block subsequent plugins/pipeline\n```\n\nUsing only `prevent_default()` still allows the LLM to generate a response.\n\n### 8. get_plugin_storage / set_plugin_storage may throw KeyError: 'owner'\n\nThis is a version mismatch between the SDK and host. Wrap storage calls in try/except:\n\n```python\ntry:\n data = await self.get_plugin_storage(\"my_key\")\nexcept Exception:\n data = None # Fallback gracefully\n```\n\n### 9. Component YAML must have full structure, not just name/description\n\n```yaml\n# ❌ WRONG — will silently fail to register the component\nname: translator\ndescription:\n en_US: 'Does stuff'\n\n# ✅ CORRECT — full component YAML\napiVersion: v1\nkind: EventListener\nmetadata:\n name: translator\n label:\n en_US: Translator\nspec:\nexecution:\n python:\n path: translator.py\n attr: Translator\n```\n\n### 10. BasePlugin import path\n\n```python\n# ❌ WRONG\nfrom langbot_plugin.api.definition.base_plugin import BasePlugin\n\n# ✅ CORRECT\nfrom langbot_plugin.api.definition.plugin import BasePlugin\n```\n\n## Pipeline Events\n\nEvents the EventListener can hook (from most general to most specific):\n\n| Event | When |\n|---|---|\n| `GroupMessageReceived` | **Any** group message arrives (before trigger rules) |\n| `PersonMessageReceived` | **Any** private message arrives |\n| `GroupNormalMessageReceived` | Group message passes trigger rules, going to LLM |\n| `PersonNormalMessageReceived` | Private message going to LLM |\n| `GroupCommandSent` | Group message matched as command |\n| `PersonCommandSent` | Private message matched as command |\n| `NormalMessageResponded` | LLM generated a response |\n| `PromptPreProcessing` | About to build LLM context |\n\n**Key insight:** `*MessageReceived` fires for ALL messages regardless of trigger rules. `*NormalMessageReceived` only fires for messages that match the pipeline's trigger rules (e.g., @bot, prefix, random%). Use `*MessageReceived` for message collection/logging.\n\n## EventContext API\n\n```python\n@self.handler(events.GroupMessageReceived)\nasync def on_msg(event_context: context.EventContext):\n event = event_context.event\n event.launcher_id # Group ID\n event.sender_id # Sender ID\n event.message_chain # MessageChain (iterate directly)\n\n # Reply to the current conversation\n await event_context.reply(MessageChain([Plain(text=\"hello\")]))\n\n # Block default pipeline behavior\n event_context.prevent_default()\n\n # Block subsequent plugins\n event_context.prevent_postorder()\n```\n\n## Setting Up a Test Environment\n\n### Deploy via Docker (GitOps + Portainer)\n\nSee `references/test-env-setup.md` for full deployment steps.\n\nQuick summary:\n1. Create `docker-compose.yaml` in `server-deploy` repo\n2. Deploy via Portainer git repository method\n3. Set up admin account via `/api/v1/user/init` POST\n4. Configure LLM provider and model via API\n5. Copy plugin to `data/plugins/` directory\n\n### WebSocket Testing\n\nLangBot's WebUI chat uses WebSocket. Connect to test message flow:\n\n```\nws://:/api/v1/pipelines//ws/connect?session_type=group\n```\n\n- `session_type=group` for group chat simulation\n- `session_type=person` for private chat (always triggers pipeline)\n\n**Requires Origin header** to pass CORS:\n```javascript\nconst ws = new WebSocket(url, {\n headers: { Origin: 'https://your-langbot-domain' }\n});\n```\n\nSend messages:\n```json\n{\"type\": \"message\", \"message\": [{\"type\": \"Plain\", \"text\": \"hello\"}]}\n```\n\nReceive:\n- `{\"type\": \"connected\", ...}` — connection established\n- `{\"type\": \"user_message\", \"data\": {...}}` — echo of sent message\n- `{\"type\": \"response\", \"data\": {\"content\": \"...\", \"is_final\": true/false}}` — bot reply (streamed)\n\n### Group Trigger Rules\n\nGroup messages only enter the pipeline if trigger rules are met:\n\n```json\n{\n \"group-respond-rules\": {\n \"at\": true, // Respond when @bot\n \"prefix\": [\"ai\"], // Respond to messages starting with \"ai\"\n \"random\": 0.0, // Probability of responding to any message (0.0-1.0)\n \"regexp\": [] // Regex patterns\n }\n}\n```\n\nFor testing, set `random: 1.0` via PUT `/api/v1/pipelines/` to respond to all messages.\n\n**Important:** EventListener hooks like `GroupMessageReceived` fire regardless of trigger rules. Only the LLM processing (`GroupNormalMessageReceived` and beyond) requires trigger rules.\n\n### Plugin Hot-Reload\n\nThere is **no hot-reload**. After changing plugin files:\n\n```bash\ndocker restart \n# Wait ~5 seconds for plugin to re-mount\n```\n\nThe main LangBot container does NOT need restart for plugin changes — only the runtime container.\n\n## API Quick Reference\n\n### Admin Setup\n\n```bash\n# Initialize admin account (first time only)\ncurl -X POST $BASE/api/v1/user/init \\\n -H \"Content-Type: application/json\" \\\n -d '{\"user\":\"admin@test.com\",\"password\":\"test123\"}'\n\n# Login\ncurl -X POST $BASE/api/v1/user/auth \\\n -H \"Content-Type: application/json\" \\\n -d '{\"user\":\"admin@test.com\",\"password\":\"test123\"}'\n# Returns: {\"data\":{\"token\":\"eyJ...\"}}\n```\n\n### Provider & Model Setup\n\n```bash\n# Create provider\ncurl -X POST $BASE/api/v1/provider/providers \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d '{\"name\":\"MyProvider\",\"requester\":\"new-api-chat-completions\",\"base_url\":\"https://api.example.com/v1\",\"api_keys\":[\"sk-xxx\"]}'\n\n# Create LLM model\ncurl -X POST $BASE/api/v1/provider/models/llm \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d '{\"name\":\"gpt-4o-mini\",\"provider_uuid\":\"\",\"abilities\":[\"chat\",\"tool-use\"]}'\n\n# List models\ncurl $BASE/api/v1/provider/models/llm -H \"Authorization: Bearer $TOKEN\"\n```\n\n### Pipeline Config\n\n```bash\n# Get pipeline\ncurl $BASE/api/v1/pipelines -H \"Authorization: Bearer $TOKEN\"\n\n# Update pipeline (e.g., set model, modify trigger rules)\ncurl -X PUT $BASE/api/v1/pipelines/ \\\n -H \"Authorization: Bearer $TOKEN\" \\\n -H \"Content-Type: application/json\" \\\n -d ''\n```\n\n## Plugin Config Types\n\nSupported `type` values in `manifest.yaml` `spec.config`:\n\n| Type | Description | Value |\n|---|---|---|\n| `string` | Text input | string |\n| `int` / `integer` | Number input | int |\n| `float` | Decimal input | float |\n| `bool` / `boolean` | Toggle | bool |\n| `select` | Dropdown (needs `options`) | string |\n| `prompt-editor` | Multi-line prompt editor | string |\n| `llm-model-selector` | LLM model picker UI | UUID string |\n| `bot-selector` | Bot picker UI | UUID string |\n\nExample — let users choose which model the plugin uses:\n\n```yaml\nspec:\n config:\n - name: model\n type: llm-model-selector\n label:\n en_US: 'LLM Model'\n zh_Hans: 'LLM 模型'\n description:\n en_US: 'Select the LLM model. Falls back to first available if not set.'\n zh_Hans: '选择 LLM 模型。未设置时使用第一个可用模型。'\n required: false\n```\n\nRead config in plugin code:\n\n```python\nmodel_uuid = self.get_config().get(\"model\")\n```\n\n## Container Restart Timing\n\nAfter plugin file changes, **only the runtime container needs restart**:\n\n```bash\ndocker restart langbot-test-runtime\n# Wait ~15 seconds before testing\n```\n\n**When to restart both (runtime first, then host):**\n- Added/removed Command or Tool components (host caches component lists)\n- Changed `manifest.yaml` structure\n\n```bash\ndocker restart langbot-test-runtime\nsleep 8\ndocker restart langbot-test\nsleep 8\n```\n\n**⚠️ Do NOT restart both simultaneously** — the host may connect before plugins are mounted, causing 502 errors or missing plugin registrations.\n\n## Debugging Checklist\n\nWhen a plugin doesn't work:\n\n1. **Check runtime logs**: `docker logs ` — look for mount/init errors\n2. **Check host logs**: `docker logs ` — look for pipeline processing errors\n3. **Verify plugin loaded**: `GET /api/v1/plugins` — should list your plugin\n4. **Test person mode first**: `session_type=person` always triggers pipeline, isolating trigger rule issues\n5. **Check trigger rules**: Group mode requires @bot, prefix match, or random% to enter pipeline\n6. **Verify model configured**: Pipeline's `config.ai.runner_config[config.ai.runner.id].model.primary` must point to a valid model UUID with working API keys\n\n## Publishing Plugins\n\nAfter testing, publish via `lbp publish`:\n\n```bash\ncd /path/to/MyPlugin\nlbp publish\n```\n\nThis builds `.lbpkg` and uploads to Space marketplace as a draft. Then go to https://space.langbot.app/market to upload screenshots and submit for review.\n\n**Prerequisite:** Must be logged in via `lbp login --token lbpat_xxx` (PAT from Space profile page).\n\n## Reference: EventListener-Only Plugin Pattern\n\nFor plugins that react to messages without commands or tools (e.g., auto-summarize URLs, collect messages, translate):\n\n```\nMyPlugin/\n├── manifest.yaml # Only EventListener in spec.components\n├── main.py # BasePlugin with shared logic (fetch, LLM calls)\n├── components/\n│ └── event_listener/\n│ ├── detector.yaml\n│ └── detector.py\n└── requirements.txt\n```\n\n**manifest.yaml** — only declare EventListener:\n```yaml\nspec:\n components:\n EventListener:\n fromDirs:\n - path: components/event_listener/\n```\n\n**detector.py** — hook `*MessageReceived`, extract text, process, reply:\n```python\n@self.handler(events.PersonMessageReceived)\nasync def on_msg(event_context: context.EventContext):\n event = event_context.event\n text_parts = []\n for component in event.message_chain:\n if isinstance(component, platform_message.Plain):\n text_parts.append(component.text)\n text = \"\".join(text_parts).strip()\n \n if should_handle(text):\n event_context.prevent_default()\n event_context.prevent_postorder()\n result = await self.plugin.process(text)\n await event_context.reply(platform_message.MessageChain([\n platform_message.Plain(text=result)\n ]))\n```\n\n**Key:** Access shared plugin logic via `self.plugin` (the BasePlugin instance).\n","frontmatter":{"name":"langbot-plugin-dev","description":"Develop, debug, and test LangBot plugins. Use when creating new LangBot plugins, fixing plugin bugs, setting up a LangBot test environment, or testing plugins via WebSocket. Covers plugin component architecture (EventListener, Command, Tool), the plugin SDK API (invoke_llm, get_llm_models, send_message, plugin storage), common pitfalls, and automated WebSocket-based testing. Triggers on \"langbot plugin\", \"lbp\", \"GroupChatSummary\", \"plugin debug\", \"langbot test\"."},"isInternal":false,"tokens":3914,"sizeBytes":15837},{"name":"curation-workflow.md","path":"skills/skills/langbot-skills-maintenance/references/curation-workflow.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-skills-maintenance/references/curation-workflow.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Curation Workflow\n\nUse this checklist when the repository starts accumulating repeated cases, copied steps, or overlapping troubleshooting entries.\n\n## Audit Pass\n\n1. Inspect the current surface:\n - `bin/lbs case list`\n - `bin/lbs case list --json --priority p0 --automation`\n - `bin/lbs case list --ready`\n - `bin/lbs case list --machine-ready`\n - `bin/lbs suite list`\n - `bin/lbs fixture list`\n - `rg \"sandbox|provider|pipeline|plugin|knowledge|mcp\" skills`\n - `rg \"If .* fails|Known Pitfalls|Debug Chat|/api/v1\" skills`\n2. Group nearby assets by intent, not by file path:\n - user-facing scenario\n - backend or provider dependency\n - failure signature\n - pass/fail evidence\n3. Pick one canonical owner:\n - stable procedures belong in `references/`\n - deterministic files and packages belong in `fixtures/` plus `fixtures/fixtures.json`\n - repeated failure signatures belong in `troubleshooting/`\n - runnable QA paths belong in `cases/`\n - reusable groups of QA paths belong in `suites/`\n - skill entry points belong in `SKILL.md`\n\n## Merge Or Split\n\nMerge when two files share the same trigger, root cause, and fix. Keep the stronger id and move missing patterns into it.\n\nSplit when a file mixes unrelated failure modes or requires different fixes. Each troubleshooting id should map to one diagnosis path.\n\nMove repeated step lists out of cases and into a reference when more than one case would need the same prompt, UI path, or log interpretation.\n\nAdd or update a suite when developers repeatedly run the same ordered group of cases. Do not copy case steps into suites; use `bin/lbs suite plan ` to expand the group.\nUse `bin/lbs suite start ` and `bin/lbs suite report --evidence-dir ` when validating that a suite is operational end to end.\n\nAdd or update `fixtures/fixtures.json` when a case depends on a deterministic file, plugin package, or local test server. The manifest should use repo-relative paths under the owning skill and should not contain machine-local absolute paths.\n\nWhen adding Debug Chat Playwright automation, reuse `scripts/e2e/lib/debug-chat.mjs` for navigation, prompt send, response leaf matching, and known failure classification. Keep case-specific prompts and expected sentinels in case YAML automation fields when possible.\n\n## Case Review\n\nFor every changed case:\n\n1. Ensure `steps` describe what to execute, not every command in the underlying implementation.\n2. Ensure `checks` contain observable UI, log, network, or filesystem evidence.\n3. Ensure `diagnostics` are fallback investigation hints, not pass criteria.\n4. Ensure `priority`, `risk`, `ci_eligible`, and `evidence_required` match the actual repeatability and evidence burden.\n5. Put must-have env vars in `env` / `automation_env`; put one-of choices such as URL-or-name in `env_any` / `automation_env_any`.\n6. Ensure linked `skills` and `troubleshooting` ids exist.\n7. Run:\n\n ```bash\n bin/lbs validate\n bin/lbs index --check\n bin/lbs index\n bin/lbs test plan \n ```\n\n## Final Gate\n\nBefore handing off:\n\n- `git diff --stat` should show a focused change set.\n- `skills.index.json` should be regenerated only by `bin/lbs index`.\n- No new asset should contain local credentials, OAuth tokens, API keys, or copied localStorage values.\n- The final note should say which checks ran and which cases or troubleshooting ids changed.\n","isInternal":false,"tokens":790,"sizeBytes":3419},{"name":"SKILL.md","path":"skills/skills/langbot-skills-maintenance/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-skills-maintenance/SKILL.md","title":"langbot-skills-maintenance","category":"anthropic-skill","format":"markdown","content":"---\nname: langbot-skills-maintenance\ndescription: Maintain the langbot-skills repository with low duplication. Use when adding, editing, or auditing LangBot skills, references, cases, troubleshooting entries, indexes, or periodic entropy-control checks for this skills repository.\n---\n\n# LangBot Skills Maintenance\n\nUse this skill before changing reusable assets in this repository.\n\n## Workflow\n\n1. Read `AGENTS.md`, `skills/.env`, and the relevant existing skill files.\n2. Classify the change:\n - `SKILL.md` for routing and concise operating rules.\n - `references/*.md` for canonical detailed workflows.\n - `cases/*.yaml` for executable test-plan skeletons.\n - `suites/*.yaml` for reusable groups of case ids.\n - `fixtures/fixtures.json` for deterministic fixture readiness metadata.\n - `reports/evidence//automation-result.json` as automation output and `reports/evidence//result.json` as final judgment output; neither is a catalog asset to commit.\n - `troubleshooting/*.yaml` for one reusable failure mode.\n3. Search existing assets before adding new files:\n - `rg \"\" skills`\n - `bin/lbs case list`\n - `bin/lbs suite list`\n - `bin/lbs fixture list`\n4. Put detail in one canonical place and link to it from cases or routing bullets.\n5. Run the checks in `AGENTS.md` after edits.\n\n## Entropy Rules\n\n- Prefer extending an existing reference or troubleshooting entry when the root cause is the same.\n- Keep cases short: setup, action, evidence, pass/fail checks. Do not paste long prompts or debug transcripts when a reference exists.\n- Put machine-checkable inputs in `env`, `automation_env`, or fixtures; put operator-confirmed assumptions in `preconditions` so `test plan` can surface `manual_check`.\n- Keep suites short: title, intent, tags, and ordered case ids. Do not duplicate case steps inside a suite.\n- Keep fixture manifests factual: id, title, path, kind, and related case ids. Do not encode environment-specific absolute paths.\n- Keep troubleshooting entries narrow: symptoms, patterns, likely causes, fixes, related assets.\n- Do not hardcode local ports, browser profile paths, secrets, tokens, or provider keys.\n- Use `bin/lbs index --check` to verify the committed index is current without writing it; run `bin/lbs index` when the index needs regeneration.\n\nFor periodic repository audits, read `references/curation-workflow.md`.\n","frontmatter":{"name":"langbot-skills-maintenance","description":"Maintain the langbot-skills repository with low duplication. Use when adding, editing, or auditing LangBot skills, references, cases, troubleshooting entries, indexes, or periodic entropy-control checks for this skills repository."},"isInternal":false,"tokens":536,"sizeBytes":2409},{"name":"SKILL.md","path":"skills/skills/langbot-space-ops/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-space-ops/SKILL.md","title":"langbot-space-ops","category":"anthropic-skill","format":"markdown","content":"---\nname: langbot-space-ops\ndescription: Browse and search the LangBot Space marketplaces (plugins, MCP servers, skills) through the Space MCP server. Use when an AI agent needs to discover LangBot extensions on space.langbot.app over MCP. Covers the /mcp endpoint, Personal Access Token (PAT) auth, the tool surface, and client configuration. Triggers on \"langbot space mcp\", \"search langbot plugins\", \"langbot marketplace mcp\", \"space.langbot.app mcp\".\n---\n\n# LangBot Space MCP Operations\n\nLangBot Space (space.langbot.app) exposes an **MCP server** so user-facing AI\nagents can browse and search the marketplaces (plugins, MCP servers, skills) and\nrank live models for automated setup.\n\n## Endpoint\n\n```\nhttps://space.langbot.app/mcp\n```\n\nTransport: **streamable HTTP** (stateless, JSON responses). For a self-hosted\nSpace instance: `http://:8383/mcp`.\n\n## Authentication\n\nReuses the existing **Personal Access Token (PAT)** — the same token the `lbp`\nCLI uses. Create one in your Space account (Profile → Personal Access Tokens),\nthen send it as a Bearer token:\n\n```\nAuthorization: Bearer \n```\n\nRequests without a valid PAT get `401 Unauthorized`.\n\n## Client configuration\n\n```json\n{\n \"mcpServers\": {\n \"langbot-space\": {\n \"url\": \"https://space.langbot.app/mcp\",\n \"headers\": { \"Authorization\": \"Bearer \" }\n }\n }\n}\n```\n\n## Tool surface\n\n| Tool | Purpose |\n| --- | --- |\n| `list_plugins` / `search_plugins` / `get_plugin` | Plugin marketplace |\n| `list_mcp_servers` / `search_mcp_servers` / `get_mcp_server` | MCP-server marketplace |\n| `list_skills` / `search_skills` / `get_skill` | Skill marketplace |\n| `select_models` | Live best-first model list for setup wizards; optional `category` filter |\n\n`list_*` and `search_*` are paged (`page`, `page_size`). `get_*` takes\n`author` + `name`. The tool surface mirrors the REST endpoints under\n`/api/v1/marketplace/*`; `select_models` mirrors `/api/v1/models/selection`.\nAll tools are read-only.\n\n## How to use\n\n1. Create a PAT in your Space account settings.\n2. Point your MCP client at `https://space.langbot.app/mcp` with the Bearer PAT.\n3. Use `search_plugins` / `search_mcp_servers` / `search_skills` to find items,\n then `get_*` for details (e.g. to obtain author/name for installation in\n LangBot itself).\n4. For automatic local-agent setup, call `select_models` (optionally with\n `category`) and choose the first compatible item. Ordering is latest probe\n state (available, unprobed, unavailable), then Space recommendation. Each\n item includes `availability.up`, `last_probed_at`, latency, and HTTP status.\n\n## Runner usage recommendations\n\nUse `search_plugins` with `runner_usage: \"agent\"` for Agent, pipeline, and\nsetup-wizard recommendations, or `runner_usage: \"event\"` for event processors.\nThe component kind remains `Runner`. Only these two exact values are accepted;\nomit the optional field to preserve unfiltered browsing.\n\n```json\n{\"query\":\"\", \"runner_usage\":\"agent\", \"page\":1, \"page_size\":100}\n```\n\nPlugin results include `latest_version` and `runner_usages: string[]`, the\nexplicit union of usages in that latest installable version. Only recommend a\nplugin when this array explicitly contains the target usage. Missing, empty,\nmalformed, or unknown usages must never mean agent-compatible. Event-only\nplugins must never enter Agent recommendations. Empty filtered results are\nvalid while legacy packages await corrected releases; never remove the filter\nto fill a recommendation list.\n\nREST callers use `runner_usage` on both\n`POST /api/v1/marketplace/extensions/search` and the compatibility\n`POST /api/v1/marketplace/plugins/search`; preserve it during fallback. Add\n`\"type_filter\":\"plugin\", \"component_filter\":\"Runner\"` on the unified endpoint.\nUsage is ANDed with other filters before pagination and `total`; MCP/Skill items\ndo not match. Invalid REST values return HTTP 400.\n\nOpen the same filter in the webpage:\n`https://space.langbot.app/market?type=plugin&component=Runner&runner_usage=agent`\n(or `runner_usage=event`). Switch All / Agent / Event in the Runner usage row.\n\n## Implementation & maintenance (for Space developers)\n\n- Server: `internal/controller/mcp/server.go` (official Go MCP SDK\n `github.com/modelcontextprotocol/go-sdk`). Tools call the service layer\n (`PluginService`, `MCPService`, `SkillService`, `ModelStatusService`) directly.\n- Mount: `internal/controller/api.go` at `/mcp` and `/mcp/*any`.\n- Auth: PAT via `AccountService.ValidatePersonalAccessToken`.\n- Docs: `docs/MCP_SERVER.md`.\n\n> When you add, remove, or change a marketplace API endpoint that should be\n> agent-accessible, update the corresponding MCP tool **and** this skill. The\n> MCP tool surface and the API must stay aligned (see `AGENTS.md`).\n\n## Pitfalls\n\n- The PAT prefix is `lbpat_` (Space), distinct from LangBot's `lbk_` API keys.\n- This server is read/browse only; it does not publish or modify marketplace\n items. Use the web UI or REST API (with appropriate auth) for that.\n","frontmatter":{"name":"langbot-space-ops","description":"Browse and search the LangBot Space marketplaces (plugins, MCP servers, skills) through the Space MCP server. Use when an AI agent needs to discover LangBot extensions on space.langbot.app over MCP. Covers the /mcp endpoint, Personal Access Token (PAT) auth, the tool surface, and client configuration. Triggers on \"langbot space mcp\", \"search langbot plugins\", \"langbot marketplace mcp\", \"space.langbot.app mcp\"."},"isInternal":false,"tokens":1219,"sizeBytes":5007},{"name":"acp-agent-runner-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/acp-agent-runner-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/acp-agent-runner-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: acp-agent-runner-debug-chat\ntitle: \"ACP Runner can answer through Debug Chat using real remote Claude\"\nmode: agent-browser\narea: pipeline\ntype: regression\npriority: p2\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - acp\n - claude\n - external-runner\n - pipeline\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nenv_any:\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/pipeline-debug-chat.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\n - LANGBOT_ACP_RUNNER_PIPELINE_URL\n - LANGBOT_ACP_RUNNER_PIPELINE_NAME\n - LANGBOT_E2E_PROMPT\n - LANGBOT_E2E_EXPECTED_TEXT\n - LANGBOT_E2E_EXPECTED_RUNNER_ID\n - LANGBOT_E2E_RESPONSE_TIMEOUT_MS\nautomation_pipeline_url_env: LANGBOT_ACP_RUNNER_PIPELINE_URL\nautomation_pipeline_name_env: LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation_expected_runner_id: \"plugin:langbot-team/ACPRunner/default\"\nautomation_prompt: \"Do not launch any background agent, subagent, task, or worker. In this current ACP session, directly call the MCP tool named langbot_get_current_event exactly once and wait for its result. After it returns, reply exactly ACP_RUNNER_E2E_OK with no other text.\"\nautomation_expected_text: \"ACP_RUNNER_E2E_OK\"\nautomation_response_timeout_ms: \"300000\"\nsetup_automation:\n - \"node:scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env\"\nsetup_provides_env:\n - LANGBOT_ACP_RUNNER_PIPELINE_URL\n - LANGBOT_ACP_RUNNER_PIPELINE_NAME\npreconditions:\n - \"The remote machine has a working Claude Code login and can run npx -y @agentclientprotocol/claude-agent-acp.\"\n - \"LangBot can non-interactively SSH to the remote machine; the runner opens the MCP reverse tunnel automatically.\"\nsteps:\n - \"Open LANGBOT_FRONTEND_URL.\"\n - \"Open the ACP Runner QA pipeline.\"\n - \"Confirm the pipeline AI runner is plugin:langbot-team/ACPRunner/default.\"\n - \"Open Debug Chat.\"\n - \"Ask the real remote Claude ACP agent to call langbot_get_current_event and return ACP_RUNNER_E2E_OK exactly.\"\nchecks:\n - \"UI: Debug Chat shows the user prompt.\"\n - \"UI: Debug Chat shows a Bot response containing ACP_RUNNER_E2E_OK.\"\n - \"Logs: Backend logs include Processing request from person_websocket and Streaming completed for this run.\"\n - \"Logs: No acp runner request error appears for this run.\"\n - \"Console: No unexpected frontend errors appear during Debug Chat.\"\nevidence_required:\n - ui\n - console\n - backend_log\ndiagnostics:\n - \"Use scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env to create/update the pipeline.\"\n - \"For remote Claude on 101, verify ssh yhh@101.34.71.12 can run without password prompts; no separate ssh -R process is required.\"\nsuccess_patterns:\n - \"ACP_RUNNER_E2E_OK\"\n - \"Processing request from person_websocket\"\n - \"Streaming completed\"\nfailure_patterns:\n - \"acp.command_not_found\"\n - \"acp.process_exited\"\n - \"Agent runner plugin:langbot-team/ACPRunner/default execution failed\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - proxy-env-mismatch\n","isInternal":false,"tokens":837,"sizeBytes":3143},{"name":"agent-run-ledger-audit.yaml","path":"skills/skills/langbot-testing/cases/agent-run-ledger-audit.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-run-ledger-audit.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-run-ledger-audit\ntitle: \"Persisted Runner run ledger passes end-to-end invariants\"\nmode: probe\narea: agent\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - ledger\n - audit\n - tools\nskills:\n - langbot-testing\nautomation: scripts/e2e/agent-run-ledger-audit.mjs\nsteps:\n - \"Set LANGBOT_AGENT_RUN_ID to audit a specific run, or leave it unset to audit the latest persisted Runner run.\"\n - \"For an external runner's own CLI tools, set LANGBOT_AGENT_TOOL_AUTHORIZATION_MODE=runner-native; keep the default strict mode for Host tool calls.\"\n - \"Read the active LangBot database configuration and inspect the selected run and its ordered events.\"\n - \"Verify completed terminal state, run.completed, paired tool.call.started/completed events, stable tool names, and monotonic ordering.\"\n - \"Compare called tools with the authorization snapshot and validate each advertised tool has owner/source, description, and parameter schema.\"\n - \"Reject malformed persisted event JSON, unrecovered invalid tool arguments, timeout, forbidden, permission-denied, or unauthorized signals.\"\n - \"Record a malformed model tool-argument payload as a recovery warning only when a later tool call succeeds and the run completes.\"\nchecks:\n - \"ledger-audit.json status is pass.\"\n - \"metrics has equal tool_call_started and tool_call_completed counts.\"\n - \"failures is empty and no unauthorized tool call is present.\"\nevidence_required:\n - filesystem\n - metrics\n - api_diagnostic\ndiagnostics:\n - \"Run this immediately after a complex UI Agent task so latest-run selection cannot drift to unrelated traffic.\"\n - \"Use LANGBOT_AGENT_RUN_ID when multiple operators share the test instance.\"\n - \"runner-native mode still checks tool pairing, stable names, ordering, terminal state, and error signals; the external runner's permission system owns its local tools.\"\nsuccess_patterns:\n - \"Agent run ledger audit passed\"\nfailure_patterns:\n - \"tool_call_pairing\"\n - \"forbidden_error_signals\"\n - \"unauthorized_tool_calls\"\ntroubleshooting:\n - box-runtime-silent-action-timeout\n","isInternal":false,"tokens":475,"sizeBytes":2104},{"name":"agent-runner-async-db-readiness.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-async-db-readiness.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-async-db-readiness.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-async-db-readiness\ntitle: \"Runner async DB readiness probe\"\nmode: probe\narea: release\ntype: smoke\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - async-db\n - aiosqlite\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-async-db-readiness.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-async-db-readiness --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation checks whether a direct aiosqlite in-memory connection can create a table within the readiness timeout.\"\nchecks:\n - \"automation-result.json status is pass or env_issue.\"\n - \"pass means async SQLite tests are worth running.\"\n - \"env_issue means async SQLite-dependent Host pytest probes should be classified as environment-limited until fixed.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"If this probe returns env_issue, run agent-runner-ledger-invariants for fast ledger coverage and skip async ledger pytest as a release blocker in this environment.\"\nsuccess_patterns:\n - \"AIOSQLITE_READY\"\nfailure_patterns:\n - \"aiosqlite readiness timed out\"\ntroubleshooting:\n - aiosqlite-connect-hangs\n","isInternal":false,"tokens":297,"sizeBytes":1202},{"name":"agent-runner-behavior-matrix.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-behavior-matrix.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-behavior-matrix.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-behavior-matrix\ntitle: \"Runner deterministic behavior matrix probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - deterministic-runner\n - protocol\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-behavior-matrix.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-behavior-matrix --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation reads fixtures/agent-runner/qa-runner-behaviors.json and validates each result sequence through the Host AgentResultNormalizer.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains QA_RUNNER_BEHAVIOR_MATRIX_OK.\"\n - \"The matrix covers ok, stream_ok, empty_output, malformed_result, and controlled_failure.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"fail means the deterministic behavior fixture and Host result normalization disagree.\"\nsuccess_patterns:\n - \"QA_RUNNER_BEHAVIOR_MATRIX_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"RunnerProtocolError\"\n - \"behavior matrix exited\"\n","isInternal":false,"tokens":279,"sizeBytes":1142},{"name":"agent-runner-fixture-contract.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-fixture-contract.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-fixture-contract.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-fixture-contract\ntitle: \"QA Runner fixture contract probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - fixture\n - deterministic-runner\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-fixture-contract.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-fixture-contract --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation imports the QA Runner fixture source and executes normal, streaming, and controlled-failure paths with SDK entities.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains QA_RUNNER_FIXTURE_CONTRACT_OK.\"\n - \"Normal input returns QA_RUNNER_OK:.\"\n - \"Streaming input emits message.delta chunks and completes.\"\n - \"Failure input returns QA_RUNNER_CONTROLLED_FAILURE.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This validates the deterministic fixture source contract. It does not prove the plugin package is installed in a live LangBot instance.\"\nsuccess_patterns:\n - \"QA_RUNNER_FIXTURE_CONTRACT_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"fixture contract exited\"\n","isInternal":false,"tokens":295,"sizeBytes":1226},{"name":"agent-runner-health-visibility.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-health-visibility.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-health-visibility.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-health-visibility\ntitle: \"Agent configuration shows whether its selected runner is usable\"\nmode: agent-browser\narea: agent\ntype: feature\npriority: p0\nrisk: medium\nci_eligible: false\ntags:\n - agent-runner\n - agent\n - productization\n - release-gate\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\nautomation: scripts/e2e/agent-runner-health-visibility.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\npreconditions:\n - \"The plugin runtime is enabled and connected, with at least one Runner registered.\"\n - \"The target is a local test instance where a temporary Agent may be created and deleted.\"\nsteps:\n - \"Read the live plugin runtime status and select a registered runner from Agent metadata.\"\n - \"Create a temporary Agent using that runner and open its Runner settings in the WebUI.\"\n - \"Confirm the UI reports that the runner is ready.\"\n - \"Replace the temporary Agent binding with a syntactically valid but unregistered runner id.\"\n - \"Reload Runner settings and confirm the UI reports that the selected runner is unavailable.\"\nchecks:\n - \"UI: A registered runner and connected plugin runtime produce a visible ready status.\"\n - \"UI: A stale runner binding produces a visible unavailable status with recovery actions.\"\n - \"API: Plugin runtime status is connected and Agent metadata contains the registered runner.\"\n - \"Console: No unexpected frontend errors appear during the flow.\"\n - \"Cleanup: The temporary Agent is deleted after evidence is collected.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - api_diagnostic\ndiagnostics:\n - \"A disconnected plugin runtime is env_issue, not a passing unavailable-runner product check.\"\n - \"A ready status inferred only from API metadata without opening the Runner UI is not a pass.\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - proxy-env-mismatch\n","isInternal":false,"tokens":470,"sizeBytes":2031},{"name":"agent-runner-ledger-concurrency.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-ledger-concurrency.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-ledger-concurrency.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-ledger-concurrency\ntitle: \"Runner run ledger concurrency and auth pytest probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - pytest\n - run-ledger\n - concurrency\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-concurrency.mjs\npreconditions:\n - \"This Host pytest probe can be slow in the current multi-repo dev environment; keep it in the release gate, but do not treat a timeout as a browser E2E failure without checking pytest logs.\"\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-concurrency --dry-run` first; remove `--dry-run` only after `agent-runner-async-db-readiness` is pass.\"\n - \"Automation resolves LANGBOT_REPO, defaulting to ../LangBot when the env var is unset.\"\n - \"Automation runs selected high-value tests from test_run_ledger_store.py and test_run_ledger_api_auth.py.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"pytest exit status is 0 for selected run ledger claim, lease, status, token, ownership, and active-claim tests.\"\n - \"pytest-stdout.log and pytest-stderr.log are written under LBS_EVIDENCE_DIR.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"env_issue means LANGBOT_REPO/default ../LangBot did not resolve, rtk/uv was unavailable, or the expected test files are missing.\"\n - \"fail means one of the selected LangBot run ledger pytest targets failed or timed out.\"\nsuccess_patterns:\n - \"pytest passed\"\nfailure_patterns:\n - \"pytest exited with status\"\n - \"pytest timed out\"\n - \"Failed to start pytest command\"\ntroubleshooting:\n - aiosqlite-connect-hangs\n","isInternal":false,"tokens":418,"sizeBytes":1663},{"name":"agent-runner-ledger-contention.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-ledger-contention.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-ledger-contention.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-ledger-contention\ntitle: \"Runner ledger SQLite contention probe\"\nmode: probe\narea: release\ntype: regression\npriority: p1\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - stress\n - ledger\n - concurrency\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-contention.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-contention --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation creates 120 queued runs in a file-backed SQLite database and uses eight worker threads to claim runs under write contention.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_CONTENTION_OK.\"\n - \"Every run is claimed once, reaches completed status, and has dispatch_attempts = 1.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe catches obvious exactly-once claim regressions under local SQLite contention; it does not replace async Host pytest or PostgreSQL concurrency checks.\"\nsuccess_patterns:\n - \"LEDGER_CONTENTION_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger contention exited\"\n","isInternal":false,"tokens":279,"sizeBytes":1170},{"name":"agent-runner-ledger-invariants.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-ledger-invariants.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-ledger-invariants.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-ledger-invariants\ntitle: \"Runner ledger schema and status invariants probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - ledger\n - invariant\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-invariants.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-invariants --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation resolves LANGBOT_REPO, defaulting to the parent LangBot checkout, and imports the sibling SDK from LANGBOT_PLUGIN_SDK_REPO or ../../langbot-plugin-sdk/src.\"\n - \"Automation checks run status sets, terminal status validation, ledger table/index DDL, and a minimal synchronous insert/read path.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_INVARIANTS_OK.\"\n - \"The probe does not require aiosqlite or browser UI.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"env_issue means LANGBOT_REPO/default ../LangBot did not resolve or Python dependencies are unavailable.\"\n - \"fail means a ledger schema/status invariant changed and the release gate needs review.\"\nsuccess_patterns:\n - \"LEDGER_INVARIANTS_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger invariant probe exited\"\n","isInternal":false,"tokens":323,"sizeBytes":1345},{"name":"agent-runner-ledger-stress.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-ledger-stress.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-ledger-stress.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-ledger-stress\ntitle: \"Runner ledger lightweight stress probe\"\nmode: probe\narea: release\ntype: regression\npriority: p1\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - stress\n - ledger\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-ledger-stress.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-ledger-stress --dry-run` first; remove `--dry-run` after checking the planned evidence directory.\"\n - \"Automation creates 100 queued runs in synchronous SQLite and simulates five runtimes claiming them in priority order.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"probe-stdout.log contains LEDGER_STRESS_OK.\"\n - \"Every run is claimed once and reaches a terminal completed status.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe is a fast deterministic stress baseline; it does not replace PostgreSQL/async concurrency tests.\"\nsuccess_patterns:\n - \"LEDGER_STRESS_OK\"\nfailure_patterns:\n - \"AssertionError\"\n - \"ledger stress exited\"\n","isInternal":false,"tokens":258,"sizeBytes":1043},{"name":"agent-runner-live-install.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-live-install.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-live-install.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-live-install\ntitle: \"QA Runner package installs and registers in LangBot\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - plugin\n - local-install\n - fixture\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_BACKEND_URL\n - LANGBOT_REPO\n - LANGBOT_E2E_LOGIN_USER\nautomation: scripts/e2e/install-qa-plugin-smoke.mjs\nautomation_plugin_package: \"skills/langbot-testing/fixtures/plugins/qa-agent-runner/dist/qa-agent-runner-0.1.0.lbpkg\"\nautomation_expected_plugin_id: \"qa/agent-runner\"\nautomation_expected_tool: \"\"\nautomation_expected_runner_id: \"plugin:qa/agent-runner/default\"\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-live-install --dry-run` first; remove `--dry-run` only after readiness points at a local test LangBot instance.\"\n - \"Automation authenticates the local test user, uploads the QA Runner .lbpkg package, waits for the install task, and reads pipeline metadata.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"/api/v1/plugins lists qa/agent-runner after install.\"\n - \"/api/v1/pipelines/_/metadata lists plugin:qa/agent-runner/default as an available runner.\"\nevidence_required:\n - api_diagnostic\n - filesystem\ndiagnostics:\n - \"This proves the deterministic package installs and registers a runner. It does not prove Debug Chat execution; use a later browser case for that.\"\n - \"If installation fails during dependencies, inspect plugin runtime logs and plugin-dependency-install-offline.\"\nsuccess_patterns:\n - \"qa/agent-runner is installed.\"\nfailure_patterns:\n - \"Plugin install task did not complete successfully\"\n - \"plugin:qa/agent-runner/default is not listed\"\ntroubleshooting:\n - plugin-runtime-timeout\n - plugin-dependency-install-offline\n","isInternal":false,"tokens":437,"sizeBytes":1790},{"name":"agent-runner-qa-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-qa-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-qa-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-qa-debug-chat\ntitle: \"QA Runner returns deterministic output through Debug Chat\"\nmode: agent-browser\narea: pipeline\ntype: regression\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - pipeline\n - debug-chat\n - fixture\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/pipeline-debug-chat.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation_pipeline_url_env: LANGBOT_QA_RUNNER_PIPELINE_URL\nautomation_pipeline_name_env: LANGBOT_QA_RUNNER_PIPELINE_NAME\nautomation_expected_runner_id: \"plugin:qa/agent-runner/default\"\nautomation_prompt: \"hello-live\"\nautomation_expected_text: \"QA_RUNNER_OK:hello-live\"\nautomation_response_timeout_ms: \"120000\"\nautomation_debug_chat_response_p95_ms: \"120000\"\nautomation_reset_debug_chat: \"1\"\nsetup_automation:\n - \"case:agent-runner-live-install\"\n - \"node:scripts/e2e/ensure-qa-agent-runner-pipeline.mjs --write-env\"\nsetup_provides_env:\n - LANGBOT_QA_RUNNER_PIPELINE_URL\n - LANGBOT_QA_RUNNER_PIPELINE_NAME\nsteps:\n - \"Open LANGBOT_FRONTEND_URL.\"\n - \"Open the pipeline from LANGBOT_QA_RUNNER_PIPELINE_URL or LANGBOT_QA_RUNNER_PIPELINE_NAME.\"\n - \"Confirm the pipeline AI runner is plugin:qa/agent-runner/default.\"\n - \"Open Debug Chat.\"\n - \"Send: hello-live.\"\nchecks:\n - \"UI: The user message appears in Debug Chat.\"\n - \"UI: A Bot message appears and contains QA_RUNNER_OK:hello-live.\"\n - \"API diagnostic: pipeline config uses plugin:qa/agent-runner/default.\"\n - \"Console: No unexpected frontend runtime errors appear during the send/receive path.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - network\n - api_diagnostic\ndiagnostics:\n - \"This is the deterministic live execution proof that sits after fixture contract and live install.\"\n - \"If the runner id mismatch is reported, rerun ensure-qa-agent-runner-pipeline.mjs --write-env.\"\nsuccess_patterns:\n - \"QA_RUNNER_OK:hello-live\"\nfailure_patterns:\n - \"plugin:qa/agent-runner/default execution failed\"\n - \"Action invoke_llm_stream call timed out\"\n - \"Agent runner temporarily unavailable\"\ntroubleshooting:\n - plugin-runtime-timeout\n","isInternal":false,"tokens":624,"sizeBytes":2386},{"name":"agent-runner-release-preflight.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-release-preflight.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-release-preflight.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-release-preflight\ntitle: \"Agent runner release gate preflight validates environment readiness\"\nmode: agent-browser\narea: release\ntype: smoke\npriority: p0\nrisk: high\nci_eligible: false\ntags:\n - agent-runner\n - release-gate\n - preflight\n - environment\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nenv_any:\n - LANGBOT_LOCAL_AGENT_PIPELINE_URL|LANGBOT_LOCAL_AGENT_PIPELINE_NAME\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\nautomation: scripts/e2e/agent-runner-release-preflight.mjs\nautomation_env:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\n - LANGBOT_BROWSER_PROFILE\n - LANGBOT_CHROMIUM_EXECUTABLE\nautomation_env_any:\n - LANGBOT_LOCAL_AGENT_PIPELINE_URL|LANGBOT_LOCAL_AGENT_PIPELINE_NAME\n - LANGBOT_ACP_RUNNER_PIPELINE_URL|LANGBOT_ACP_RUNNER_PIPELINE_NAME\npreconditions:\n - \"LANGBOT_LOCAL_AGENT_PIPELINE_URL or LANGBOT_LOCAL_AGENT_PIPELINE_NAME points to the local-agent release pipeline.\"\n - \"LANGBOT_ACP_RUNNER_PIPELINE_URL or LANGBOT_ACP_RUNNER_PIPELINE_NAME points to the ACP Runner release pipeline.\"\n - \"The active browser profile is authenticated for the same LangBot backend.\"\n - \"By default the preflight performs a cheap model test for the local-agent primary model; set LANGBOT_PREFLIGHT_TEST_MODELS=0 only when deliberately classifying model credentials outside this run.\"\nsteps:\n - \"Open LANGBOT_FRONTEND_URL with the configured browser profile.\"\n - \"Use the browser token to call LangBot backend readiness APIs without printing token values.\"\n - \"Check plugin runtime status, Box status, required runner plugins, qa-plugin-smoke, and qa_plugin_echo.\"\n - \"Resolve the local-agent and ACP Runner QA pipelines from their case-specific env vars.\"\n - \"Assert each pipeline uses the expected runner id.\"\n - \"Assert the external runner pipeline uses the expected runner id.\"\n - \"Assert the local-agent primary model advertises func_call and vision for the full release gate.\"\n - \"Run the local-agent primary model test endpoint unless LANGBOT_PREFLIGHT_TEST_MODELS=0.\"\nchecks:\n - \"API diagnostic: api-diagnostic.json has no blockers and no env_issues.\"\n - \"API diagnostic: required pipelines resolve to plugin:langbot-team/LocalAgent/default and plugin:langbot-team/ACPRunner/default.\"\n - \"API diagnostic: qa_plugin_echo is exposed by /api/v1/tools.\"\n - \"API diagnostic: local-agent model check catches invalid credentials or missing func_call/vision before release E2E starts.\"\n - \"Secret safety: token values, api keys, and provider secrets are not printed.\"\nevidence_required:\n - ui\n - screenshot\n - console\n - network\n - api_diagnostic\ndiagnostics:\n - \"blocked means the test instance is not configured for the full release gate: missing pipeline, wrong runner id, or missing plugin.\"\n - \"env_issue means the runtime or upstream dependency is not usable: backend unavailable, plugin runtime down, Box down, or the local-agent model cannot pass a model test.\"\n - \"If qa_mcp_echo is absent here, continue to mcp-stdio-register before mcp-stdio-tool-call; qa_mcp_echo is not required before registration.\"\n - \"If the model check fails with invalid api key, switch the local-agent release pipeline to a known-good func_call model before diagnosing runner code.\"\nsuccess_patterns:\n - \"Release gate preflight passed\"\nfailure_patterns:\n - \"Preflight blocked\"\n - \"Preflight environment issue\"\n - \"invalid api key\"\n - \"runner.llm_error\"\ntroubleshooting:\n - backend-not-listening\n - plugin-runtime-timeout\n - local-agent-model-route-unavailable\n - proxy-env-mismatch\n","isInternal":false,"tokens":855,"sizeBytes":3615},{"name":"agent-runner-runtime-chaos.yaml","path":"skills/skills/langbot-testing/cases/agent-runner-runtime-chaos.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/agent-runner-runtime-chaos.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"id: agent-runner-runtime-chaos\ntitle: \"Runner SDK runtime chaos pytest probe\"\nmode: probe\narea: release\ntype: regression\npriority: p0\nrisk: high\nci_eligible: true\ntags:\n - agent-runner\n - probe\n - pytest\n - runtime\n - sdk\nskills:\n - langbot-testing\nenv:\nautomation: skills/langbot-testing/probes/agent-runner-runtime-chaos.mjs\nsteps:\n - \"Run `rtk bin/lbs test run agent-runner-runtime-chaos --dry-run` first; remove `--dry-run` after checking the SDK repo target and evidence directory.\"\n - \"Automation resolves LANGBOT_PLUGIN_SDK_REPO, defaulting to ../../langbot-plugin-sdk when the env var is unset.\"\n - \"Automation runs the existing SDK pytest files tests/runtime/plugin/test_mgr_runner.py and tests/runtime/test_pull_api_handlers.py.\"\nchecks:\n - \"automation-result.json status is pass.\"\n - \"pytest exit status is 0 for the existing Runner runtime and pull API handler tests.\"\n - \"pytest-stdout.log and pytest-stderr.log are written under LBS_EVIDENCE_DIR.\"\nevidence_required:\n - filesystem\ndiagnostics:\n - \"This probe does not open the WebUI; it runs SDK pytest targets directly.\"\n - \"env_issue means LANGBOT_PLUGIN_SDK_REPO/default ../../langbot-plugin-sdk did not resolve, rtk/uv was unavailable, or the expected test files are missing.\"\n - \"fail means the existing SDK runtime pytest target failed or timed out.\"\nsuccess_patterns:\n - \"pytest passed\"\nfailure_patterns:\n - \"pytest exited with status\"\n - \"pytest timed out\"\n - \"Failed to start pytest command\"\n","isInternal":false,"tokens":364,"sizeBytes":1485},{"name":"bot-event-routing-product-flow.yaml","path":"skills/skills/langbot-testing/cases/bot-event-routing-product-flow.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/bot-event-routing-product-flow.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":666,"sizeBytes":2929},{"name":"box-mcp-heartbeat-recovery.yaml","path":"skills/skills/langbot-testing/cases/box-mcp-heartbeat-recovery.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/box-mcp-heartbeat-recovery.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1005,"sizeBytes":4276},{"name":"claude-code-agent-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/claude-code-agent-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/claude-code-agent-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":901,"sizeBytes":3686},{"name":"codex-agent-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/codex-agent-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/codex-agent-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":826,"sizeBytes":3352},{"name":"dify-agent-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/dify-agent-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/dify-agent-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":521,"sizeBytes":2136},{"name":"langbot-fake-provider-debug-chat-cross-pipeline-isolation.yaml","path":"skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-cross-pipeline-isolation.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-cross-pipeline-isolation.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":912,"sizeBytes":3934},{"name":"langbot-fake-provider-debug-chat-fault-recovery.yaml","path":"skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-fault-recovery.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-fault-recovery.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1030,"sizeBytes":4473},{"name":"langbot-fake-provider-debug-chat-load.yaml","path":"skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-load.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-load.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":818,"sizeBytes":3477},{"name":"langbot-fake-provider-debug-chat-slow-load.yaml","path":"skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-slow-load.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-fake-provider-debug-chat-slow-load.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":893,"sizeBytes":3816},{"name":"langbot-fault-taxonomy-contract.yaml","path":"skills/skills/langbot-testing/cases/langbot-fault-taxonomy-contract.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-fault-taxonomy-contract.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":368,"sizeBytes":1559},{"name":"langbot-live-backend-latency.yaml","path":"skills/skills/langbot-testing/cases/langbot-live-backend-latency.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-live-backend-latency.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":398,"sizeBytes":1625},{"name":"langbot-live-backend-log-health.yaml","path":"skills/skills/langbot-testing/cases/langbot-live-backend-log-health.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-live-backend-log-health.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":498,"sizeBytes":2013},{"name":"langbot-live-control-plane-api.yaml","path":"skills/skills/langbot-testing/cases/langbot-live-control-plane-api.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-live-control-plane-api.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":499,"sizeBytes":2042},{"name":"langbot-overhead-accounting-contract.yaml","path":"skills/skills/langbot-testing/cases/langbot-overhead-accounting-contract.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-overhead-accounting-contract.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":413,"sizeBytes":1757},{"name":"langbot-space-debug-chat-concurrency-smoke.yaml","path":"skills/skills/langbot-testing/cases/langbot-space-debug-chat-concurrency-smoke.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langbot-space-debug-chat-concurrency-smoke.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":809,"sizeBytes":3489},{"name":"langrag-kb-retrieve.yaml","path":"skills/skills/langbot-testing/cases/langrag-kb-retrieve.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langrag-kb-retrieve.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":615,"sizeBytes":2303},{"name":"langrag-parser-golden-e2e.yaml","path":"skills/skills/langbot-testing/cases/langrag-parser-golden-e2e.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langrag-parser-golden-e2e.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":893,"sizeBytes":3620},{"name":"langrag-sentinel-kb-discover.yaml","path":"skills/skills/langbot-testing/cases/langrag-sentinel-kb-discover.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/langrag-sentinel-kb-discover.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":375,"sizeBytes":1399},{"name":"local-agent-basic-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-basic-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-basic-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":646,"sizeBytes":2628},{"name":"local-agent-combo-rag-compaction-tool-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-combo-rag-compaction-tool-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-combo-rag-compaction-tool-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1853,"sizeBytes":7553},{"name":"local-agent-complex-coding-task-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-complex-coding-task-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-complex-coding-task-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1276,"sizeBytes":5540},{"name":"local-agent-context-compaction-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-context-compaction-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-context-compaction-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1796,"sizeBytes":7882},{"name":"local-agent-effective-prompt-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-effective-prompt-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-effective-prompt-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":720,"sizeBytes":3052},{"name":"local-agent-model-fallback-before-first-chunk-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-model-fallback-before-first-chunk-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-model-fallback-before-first-chunk-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":828,"sizeBytes":3585},{"name":"local-agent-multimodal-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-multimodal-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-multimodal-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":737,"sizeBytes":3113},{"name":"local-agent-multitool-rag-compaction-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-multitool-rag-compaction-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-multitool-rag-compaction-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1826,"sizeBytes":7389},{"name":"local-agent-nonstreaming-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-nonstreaming-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-nonstreaming-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":622,"sizeBytes":2584},{"name":"local-agent-parallel-tools-rag-compaction-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-parallel-tools-rag-compaction-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-parallel-tools-rag-compaction-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1808,"sizeBytes":7535},{"name":"local-agent-plugin-tool-call-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-plugin-tool-call-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-plugin-tool-call-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":899,"sizeBytes":3735},{"name":"local-agent-rag-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-rag-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-rag-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":922,"sizeBytes":3686},{"name":"local-agent-rag-multimodal-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-rag-multimodal-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-rag-multimodal-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":789,"sizeBytes":3217},{"name":"local-agent-steering-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-steering-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-steering-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":908,"sizeBytes":3816},{"name":"local-agent-streaming-post-commit-failure-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-streaming-post-commit-failure-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-streaming-post-commit-failure-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":914,"sizeBytes":3986},{"name":"local-agent-tool-error-recovery-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-tool-error-recovery-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-tool-error-recovery-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1069,"sizeBytes":4553},{"name":"local-agent-tool-loop-limit-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/local-agent-tool-loop-limit-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/local-agent-tool-loop-limit-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1003,"sizeBytes":4272},{"name":"mcp-stdio-register.yaml","path":"skills/skills/langbot-testing/cases/mcp-stdio-register.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/mcp-stdio-register.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":635,"sizeBytes":2456},{"name":"mcp-stdio-tool-call.yaml","path":"skills/skills/langbot-testing/cases/mcp-stdio-tool-call.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/mcp-stdio-tool-call.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":1243,"sizeBytes":4953},{"name":"pipeline-debug-chat-performance.yaml","path":"skills/skills/langbot-testing/cases/pipeline-debug-chat-performance.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/pipeline-debug-chat-performance.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":724,"sizeBytes":2987},{"name":"pipeline-debug-chat.yaml","path":"skills/skills/langbot-testing/cases/pipeline-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/pipeline-debug-chat.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":563,"sizeBytes":2268},{"name":"plugin-e2e-smoke.yaml","path":"skills/skills/langbot-testing/cases/plugin-e2e-smoke.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/plugin-e2e-smoke.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":706,"sizeBytes":3013},{"name":"provider-deepseek.yaml","path":"skills/skills/langbot-testing/cases/provider-deepseek.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/provider-deepseek.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":374,"sizeBytes":1552},{"name":"qa-plugin-smoke-live-install.yaml","path":"skills/skills/langbot-testing/cases/qa-plugin-smoke-live-install.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/qa-plugin-smoke-live-install.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":432,"sizeBytes":1719},{"name":"sandbox-skill-authoring-e2e.yaml","path":"skills/skills/langbot-testing/cases/sandbox-skill-authoring-e2e.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/sandbox-skill-authoring-e2e.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":347,"sizeBytes":1510},{"name":"sandbox-skill-authoring-edit-existing-e2e.yaml","path":"skills/skills/langbot-testing/cases/sandbox-skill-authoring-edit-existing-e2e.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/sandbox-skill-authoring-edit-existing-e2e.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":432,"sizeBytes":1824},{"name":"skill-discovery-via-mcp-gateway.yaml","path":"skills/skills/langbot-testing/cases/skill-discovery-via-mcp-gateway.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/skill-discovery-via-mcp-gateway.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":749,"sizeBytes":3072},{"name":"webui-login-state.yaml","path":"skills/skills/langbot-testing/cases/webui-login-state.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/webui-login-state.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":342,"sizeBytes":1329},{"name":"wizard-onebot-agent-runtime.yaml","path":"skills/skills/langbot-testing/cases/wizard-onebot-agent-runtime.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/wizard-onebot-agent-runtime.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":591,"sizeBytes":2460},{"name":"wizard-runner-marketplace-catalog.yaml","path":"skills/skills/langbot-testing/cases/wizard-runner-marketplace-catalog.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/wizard-runner-marketplace-catalog.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":558,"sizeBytes":2464},{"name":"wizard-scenario-routing.yaml","path":"skills/skills/langbot-testing/cases/wizard-scenario-routing.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/wizard-scenario-routing.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":527,"sizeBytes":2231},{"name":"workspace-compatibility-preflight.yaml","path":"skills/skills/langbot-testing/cases/workspace-compatibility-preflight.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/workspace-compatibility-preflight.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":370,"sizeBytes":1618},{"name":"workspace-plugin-pages-smoke.yaml","path":"skills/skills/langbot-testing/cases/workspace-plugin-pages-smoke.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/workspace-plugin-pages-smoke.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":488,"sizeBytes":2012},{"name":"workspace-repository-contracts.yaml","path":"skills/skills/langbot-testing/cases/workspace-repository-contracts.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/cases/workspace-repository-contracts.yaml","title":"Cases Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":335,"sizeBytes":1515},{"name":"agent-runner-qa-workflow.md","path":"skills/skills/langbot-testing/references/agent-runner-qa-workflow.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/agent-runner-qa-workflow.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":1986,"sizeBytes":8318},{"name":"agent-runner-release-gate.md","path":"skills/skills/langbot-testing/references/agent-runner-release-gate.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/agent-runner-release-gate.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":2478,"sizeBytes":10849},{"name":"dify-agent-runner.md","path":"skills/skills/langbot-testing/references/dify-agent-runner.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/dify-agent-runner.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":509,"sizeBytes":2174},{"name":"langrag-knowledge-base.md","path":"skills/skills/langbot-testing/references/langrag-knowledge-base.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/langrag-knowledge-base.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":1270,"sizeBytes":6059},{"name":"local-agent-runner-coverage.md","path":"skills/skills/langbot-testing/references/local-agent-runner-coverage.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/local-agent-runner-coverage.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":2530,"sizeBytes":12048},{"name":"local-agent-runner.md","path":"skills/skills/langbot-testing/references/local-agent-runner.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/local-agent-runner.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":1384,"sizeBytes":6570},{"name":"mcp-stdio-testing.md","path":"skills/skills/langbot-testing/references/mcp-stdio-testing.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/mcp-stdio-testing.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":979,"sizeBytes":4095},{"name":"model-provider-testing.md","path":"skills/skills/langbot-testing/references/model-provider-testing.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/model-provider-testing.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":215,"sizeBytes":951},{"name":"performance-reliability-testing.md","path":"skills/skills/langbot-testing/references/performance-reliability-testing.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/performance-reliability-testing.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":2790,"sizeBytes":12928},{"name":"pipeline-debug-chat.md","path":"skills/skills/langbot-testing/references/pipeline-debug-chat.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/pipeline-debug-chat.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":383,"sizeBytes":1750},{"name":"plugin-e2e-smoke.md","path":"skills/skills/langbot-testing/references/plugin-e2e-smoke.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/plugin-e2e-smoke.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":623,"sizeBytes":2493},{"name":"sandbox-skill-authoring.md","path":"skills/skills/langbot-testing/references/sandbox-skill-authoring.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/sandbox-skill-authoring.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":709,"sizeBytes":3405},{"name":"skill-all-tool-acceptance.md","path":"skills/skills/langbot-testing/references/skill-all-tool-acceptance.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/skill-all-tool-acceptance.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":2601,"sizeBytes":10336},{"name":"troubleshooting.md","path":"skills/skills/langbot-testing/references/troubleshooting.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/troubleshooting.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":955,"sizeBytes":4589},{"name":"web-ui-testing.md","path":"skills/skills/langbot-testing/references/web-ui-testing.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/web-ui-testing.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":305,"sizeBytes":1465},{"name":"workspace-release-testing.md","path":"skills/skills/langbot-testing/references/workspace-release-testing.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/references/workspace-release-testing.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"","isInternal":false,"tokens":436,"sizeBytes":2210},{"name":"SKILL.md","path":"skills/skills/langbot-testing/SKILL.md","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/SKILL.md","title":"langbot-testing","category":"anthropic-skill","format":"markdown","content":"","frontmatter":{"name":"langbot-testing","description":"Test LangBot WebUI and core product flows with an automated browser and backend logs. Use when validating the configured LangBot frontend, pipeline Debug Chat, model provider setup and test buttons, bot and knowledge-base UI flows, or troubleshooting failed LangBot end-to-end tests."},"isInternal":false,"tokens":896,"sizeBytes":4015},{"name":"agent-runner-release-gate.yaml","path":"skills/skills/langbot-testing/suites/agent-runner-release-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/agent-runner-release-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":392,"sizeBytes":1518},{"name":"core-smoke.yaml","path":"skills/skills/langbot-testing/suites/core-smoke.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/core-smoke.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":84,"sizeBytes":313},{"name":"eba-release-gate.yaml","path":"skills/skills/langbot-testing/suites/eba-release-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/eba-release-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":114,"sizeBytes":456},{"name":"langbot-debug-chat-isolation-gate.yaml","path":"skills/skills/langbot-testing/suites/langbot-debug-chat-isolation-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/langbot-debug-chat-isolation-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":109,"sizeBytes":464},{"name":"langbot-debug-chat-load-gate.yaml","path":"skills/skills/langbot-testing/suites/langbot-debug-chat-load-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/langbot-debug-chat-load-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":164,"sizeBytes":697},{"name":"langbot-live-backend-gate.yaml","path":"skills/skills/langbot-testing/suites/langbot-live-backend-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/langbot-live-backend-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":97,"sizeBytes":413},{"name":"langbot-performance-contract-gate.yaml","path":"skills/skills/langbot-testing/suites/langbot-performance-contract-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/langbot-performance-contract-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":89,"sizeBytes":384},{"name":"langbot-performance-reliability-gate.yaml","path":"skills/skills/langbot-testing/suites/langbot-performance-reliability-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/langbot-performance-reliability-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":126,"sizeBytes":521},{"name":"langbot-user-path-performance-gate.yaml","path":"skills/skills/langbot-testing/suites/langbot-user-path-performance-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/langbot-user-path-performance-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":77,"sizeBytes":329},{"name":"langbot-workspace-contract-gate.yaml","path":"skills/skills/langbot-testing/suites/langbot-workspace-contract-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/langbot-workspace-contract-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":137,"sizeBytes":553},{"name":"langbot-workspace-release-gate.yaml","path":"skills/skills/langbot-testing/suites/langbot-workspace-release-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/langbot-workspace-release-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":216,"sizeBytes":850},{"name":"local-agent-gate.yaml","path":"skills/skills/langbot-testing/suites/local-agent-gate.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/suites/local-agent-gate.yaml","title":"Suites Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":275,"sizeBytes":1101},{"name":"agent-runner-actor-context-fields.yaml","path":"skills/skills/langbot-testing/troubleshooting/agent-runner-actor-context-fields.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/agent-runner-actor-context-fields.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":271,"sizeBytes":1183},{"name":"aiosqlite-connect-hangs.yaml","path":"skills/skills/langbot-testing/troubleshooting/aiosqlite-connect-hangs.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/aiosqlite-connect-hangs.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":337,"sizeBytes":1418},{"name":"ambiguous-runner-default-label.yaml","path":"skills/skills/langbot-testing/troubleshooting/ambiguous-runner-default-label.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/ambiguous-runner-default-label.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":282,"sizeBytes":1199},{"name":"backend-not-listening.yaml","path":"skills/skills/langbot-testing/troubleshooting/backend-not-listening.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/backend-not-listening.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":354,"sizeBytes":1453},{"name":"box-runtime-silent-action-timeout.yaml","path":"skills/skills/langbot-testing/troubleshooting/box-runtime-silent-action-timeout.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/box-runtime-silent-action-timeout.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":353,"sizeBytes":1629},{"name":"box-session-conflict-logical-metadata.yaml","path":"skills/skills/langbot-testing/troubleshooting/box-session-conflict-logical-metadata.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/box-session-conflict-logical-metadata.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":310,"sizeBytes":1415},{"name":"debug-chat-history-contaminates-automation.yaml","path":"skills/skills/langbot-testing/troubleshooting/debug-chat-history-contaminates-automation.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/debug-chat-history-contaminates-automation.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":319,"sizeBytes":1555},{"name":"dynamic-form-missing-config-id.yaml","path":"skills/skills/langbot-testing/troubleshooting/dynamic-form-missing-config-id.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/dynamic-form-missing-config-id.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":225,"sizeBytes":983},{"name":"e2b-extra-mount-sync-missing.yaml","path":"skills/skills/langbot-testing/troubleshooting/e2b-extra-mount-sync-missing.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/e2b-extra-mount-sync-missing.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":372,"sizeBytes":1573},{"name":"local-agent-mcp-text-result-empty-follow-up.yaml","path":"skills/skills/langbot-testing/troubleshooting/local-agent-mcp-text-result-empty-follow-up.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/local-agent-mcp-text-result-empty-follow-up.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":372,"sizeBytes":1668},{"name":"local-agent-model-route-unavailable.yaml","path":"skills/skills/langbot-testing/troubleshooting/local-agent-model-route-unavailable.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/local-agent-model-route-unavailable.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":637,"sizeBytes":2836},{"name":"marketplace-network-flaky.yaml","path":"skills/skills/langbot-testing/troubleshooting/marketplace-network-flaky.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/marketplace-network-flaky.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":250,"sizeBytes":1066},{"name":"mcp-stdio-args-not-applied.yaml","path":"skills/skills/langbot-testing/troubleshooting/mcp-stdio-args-not-applied.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/mcp-stdio-args-not-applied.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":583,"sizeBytes":2461},{"name":"nsjail-cli-compatibility.yaml","path":"skills/skills/langbot-testing/troubleshooting/nsjail-cli-compatibility.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/nsjail-cli-compatibility.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":447,"sizeBytes":1771},{"name":"pipeline-form-controlled-warning.yaml","path":"skills/skills/langbot-testing/troubleshooting/pipeline-form-controlled-warning.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/pipeline-form-controlled-warning.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":219,"sizeBytes":1041},{"name":"plugin-dependency-install-offline.yaml","path":"skills/skills/langbot-testing/troubleshooting/plugin-dependency-install-offline.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/plugin-dependency-install-offline.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":397,"sizeBytes":1775},{"name":"plugin-runtime-timeout.yaml","path":"skills/skills/langbot-testing/troubleshooting/plugin-runtime-timeout.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/plugin-runtime-timeout.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":371,"sizeBytes":1712},{"name":"provider-image-parse-error.yaml","path":"skills/skills/langbot-testing/troubleshooting/provider-image-parse-error.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/provider-image-parse-error.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":298,"sizeBytes":1432},{"name":"proxy-env-mismatch.yaml","path":"skills/skills/langbot-testing/troubleshooting/proxy-env-mismatch.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/proxy-env-mismatch.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":258,"sizeBytes":1104},{"name":"sandbox-native-tools-unavailable.yaml","path":"skills/skills/langbot-testing/troubleshooting/sandbox-native-tools-unavailable.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/sandbox-native-tools-unavailable.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":411,"sizeBytes":1795},{"name":"socks-proxy-without-socksio.yaml","path":"skills/skills/langbot-testing/troubleshooting/socks-proxy-without-socksio.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/socks-proxy-without-socksio.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":326,"sizeBytes":1337},{"name":"survey-widget-blocks-debug-chat.yaml","path":"skills/skills/langbot-testing/troubleshooting/survey-widget-blocks-debug-chat.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/survey-widget-blocks-debug-chat.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":234,"sizeBytes":1022},{"name":"telemetry-proxy-noise.yaml","path":"skills/skills/langbot-testing/troubleshooting/telemetry-proxy-noise.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/telemetry-proxy-noise.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":296,"sizeBytes":1424},{"name":"tool-name-collision-between-mcp-and-plugin.yaml","path":"skills/skills/langbot-testing/troubleshooting/tool-name-collision-between-mcp-and-plugin.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/tool-name-collision-between-mcp-and-plugin.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":396,"sizeBytes":1743},{"name":"uv-run-resyncs-local-sdk.yaml","path":"skills/skills/langbot-testing/troubleshooting/uv-run-resyncs-local-sdk.yaml","rawUrl":"https://raw.githubusercontent.com/langbot-app/LangBot/HEAD/skills/skills/langbot-testing/troubleshooting/uv-run-resyncs-local-sdk.yaml","title":"Troubleshooting Skill","category":"anthropic-skill","format":"yaml","content":"","isInternal":false,"tokens":343,"sizeBytes":1410}],"systemPromptSnippet":"\n\n# AGENTS.md\n\nThis file guides code agents working in the LangBot main repository. `CLAUDE.md` is a symlink to this file.\n\nRead `ARCHITECTURE.md` before non-trivial backend, frontend, runtime, plugin, Box, MCP, persistence, or cross-repo SDK changes. This file is the working checklist; `ARCHITECTURE.md` is the system map.\n\n## Quick Facts\n\n- Python backend: `>=3.11,<4.0`, dependencies managed by `uv`.\n- Frontend: `web/` is Vite + React Router 7 + shadcn/ui + Tailwind, managed by `pnpm`.\n- Backend framework: Quart served by Hypercorn on `api.port`, default `5300`.\n- Frontend dev server: `web/` on `3000`, with `VITE_API_BASE_URL` pointing at the backend.\n- Plugin/Box/runtime contracts live in sibling repo `langbot-plugin-sdk`, pinned as `langbot-plugin` in `pyproject.toml`.\n\n## Essential Commands\n\n```bash\nuv sync --dev\nuv run main.py\nuv run pre-commit install\n\ncd web\npnpm install\npnpm dev\npnpm build\n```\n\nUseful focused tests:\n\n```bash\nuv run pytest tests/unit_tests -q\nuv run pytest tests/integration -q\nuv run pytest tests/integration/persistence -q\nuv run pytest tests/manual/mcp_smoke.py\n\ncd web\npnpm lint\npnpm test:e2e\n```\n\nRun the narrowest useful test first, then broader checks when confidence is needed.\n\n## Where to Look\n\n- Architecture map: `ARCHITECTURE.md`.\n- Dev environment guide: https://langbot.app/docs/zh/develop/dev-config.\n- Plugin runtime / CLI / SDK debugging: https://langbot.app/docs/zh/develop/plugin-runtime.\n- API-key auth: `docs/API_KEY_AUTH.md`.\n- Service API CLI: [`langbot-cli`](https://github.com/langbot-app/langbot-cli) (`lbctl`) manages running Workspaces; it is separate from the SDK's `lbp` CLI.\n- Box deep-dive notes: `docs/review/box-architecture.md` and related files.\n- In-repo skills: `skills/` is the single source of truth for LangBot agent skills.\n- SDK repo: `../langbot-plugin-sdk/` when changing shared entities, plugin APIs, action protocol, `lbp rt`, or `lbp box`.\n\n## Cross-Repo SDK Work\n\nWhen changing SDK contracts used by LangBot:\n\n```bash\n# from langbot-plugin-sdk, with LangBot's .venv active\nuv pip install .\n\n# from LangBot, preserve the locally installed SDK\nuv run --no-sync main.py\n```\n\nFor standalone runtime debugging:\n\n```bash\n# in langbot-plugin-sdk\nuv run --no-sync lbp rt\nuv run --no-sync lbp box\n\n# in LangBot\nuv run --no-sync main.py --standalone-runtime\nuv run --no-sync main.py --standalone-box\n```\n\nConfig keys to verify in `data/config.yaml` / `src/langbot/templates/config.yaml`:\n\n- Plugin runtime: `plugin.runtime_ws_url`, default Docker host `langbot_plugin_runtime:5400/control/ws`.\n- Box runtime: `box.enabled`, `box.backend`, `box.runtime.endpoint`, Docker host `langbot_box:5410`.\n- API/MCP auth: `api.global_api_key`.\n\n## Change Rules\n\n- HTTP API changes that should be agent-accessible must update the matching MCP tool in `src/langbot/pkg/api/mcp/server.py` and the relevant skill under `skills/` in the same pass.\n- When changing Service API routes or capabilities used by `lbctl`, check compatibility with the separate `langbot-cli` repository.\n- New schema changes use Alembic under `src/langbot/pkg/persistence/alembic/versions/`. LangBot 4.x does not support upgrading 3.x databases.\n- New platform behavior belongs in platform adapters only for platform translation; pipeline/business logic belongs in `pkg/pipeline/` or services.\n- User-facing strings must support i18n; follow the backend language rules below for server-generated content.\n- Conversation payload columns, including new ones, must use `MediaReferenceText` and persist image references only. See the [media persistence contract](ARCHITECTURE.md#persistence-and-configuration).\n- Interactive session lists use `SessionSummaryReader` to read only `monitoring_sessions`; `SessionSummaryWriter` maintains summaries in the source transaction. Historical initialization completes before writers start, never in requests or application background tasks. Follow the [conversation summary contract](ARCHITECTURE.md#conversation-summary-boundary) and keep `uv run pytest --required-suite session_summary` passing.\n- Code comments and docstrings must be English.\n- Keep compatibility and security in mind; LangBot is used in both self-hosted/community and toB deployments.\n- Commit message format: `(): `.\n\n## Backend Message Language\n\n- LangBot-owned business errors, runtime notices, and default messages sent to users must use the owning Workspace's message language. Keep this independent of each member's WebUI language; an unset Workspace language defaults to `en_US`.\n- Reuse `src/langbot/pkg/i18n/` and the existing `message_language` entry in `WorkspaceMetadata`. Use `t()` with stable English templates and named parameters; add matching keys and placeholders to all eight catalogs: `en_US`, `zh_Hans`, `zh_Hant`, `ja_JP`, `es_ES`, `ru_RU`, `th_TH`, and `vi_VN`. Avoid new translation frameworks or duplicate language settings.\n- Resolve language from a trusted Workspace or bot execution context, and isolate it across concurrent requests, tasks, and adapter callbacks. Browser language hints may supply account-only or unauthenticated response language, but must not override a resolved Workspace or authorize tenant selection. Reading/changing the setting requires `workspace.view` / `workspace.update`.\n- Pipeline notices should be configurable in Pipeline output settings through `pipeline_notice()`. Blank overrides inherit the Workspace translation. Preserve user-authored text literally except for documented placeholders, and retain existing visibility switches. Do not rewrite historical messages or custom text when the language changes.\n- China-only adapters and configuration/field validation diagnostics do not require localization. An adapter also serving international users is not exempt merely because it supports a Chinese platform.\n- Preserve raw provider/plugin diagnostics, tracebacks, protocol values, and machine-readable error/status codes. Translate LangBot's own category or explanation around them; do not translate arbitrary errors with substring replacement. Plugin-authored messages remain the plugin's localization responsibility.\n- Keep changes small and maintainable. Reuse existing persistence and response boundaries; delete obsolete fields only after verifying they have no remaining readers, including compatibility paths.\n- When changing language behavior, verify catalog/placeholder parity, Workspace isolation and permissions, and Pipeline inheritance/custom overrides. Update affected configuration snapshots and API/MCP documentation. Implementation details and focused test locations: [Backend message language](docs/architecture/backend-i18n.md).\n\n## Runtime Pitfalls\n\n- Local stdio Plugin Runtime disconnects do not auto-reconnect; restart LangBot if that path breaks.\n- Orphan runtime processes on `5400`/`5401` commonly break plugin debugging.\n- Use `uv run --no-sync` after locally installing the SDK, or `uv` may restore the pinned package.\n- A false Box “no backend” often means Docker is running but the current user lacks Docker socket permission.\n- Do not confuse external MCP servers LangBot connects to (`pkg/provider/tools/loaders/mcp.py`) with LangBot's own `/mcp` server (`pkg/api/mcp/`).\n- `CLAUDE.md` is a symlink to this file; edit `AGENTS.md`, not the symlink.\n\n## Principles\n\n- Keep it simple, stupid.\n- Entities should not be multiplied unnecessarily.\n- 八荣八耻\n\n 以瞎猜接口为耻，以认真查询为荣。\n 以模糊执行为耻，以寻求确认为荣。\n 以臆想业务为耻，以人类确认为荣。\n 以创造接口为耻，以复用现有为荣。\n 以跳过验证为耻，以主动测试为荣。\n 以破坏架构为耻，以遵循规范为荣。\n 以假装理解为耻，以诚实无知为荣。\n 以盲目修改为耻，以谨慎重构为荣。\n\n\n# Agent Workflow\n\nThis repository stores reusable LangBot agent-testing assets. Keep changes structured so the next agent does not need to rediscover paths.\n\n## First Steps\n\n1. Read `skills/.env` before using local URLs, paths, browser profiles, or proxy defaults. If present, `skills/.env.local` overrides it for this machine and must not be committed. On a new machine, copy `skills/.env.example` to `skills/.env.local` first.\n2. Pick the smallest relevant skill:\n - `langbot-env-setup` for environment, browser, OAuth, proxy, and startup.\n - `langbot-testing` for WebUI, provider, pipeline, cases, and troubleshooting.\n - `langbot-skills-maintenance` for adding, deduplicating, or auditing this skills repository.\n3. Prefer existing cases and troubleshooting entries before exploring from scratch.\n\n## Editing Rules\n\n- UI/browser testing is the primary QA path. API/curl checks are diagnostic only and cannot make a UI case pass by themselves.\n- Put skills under `skills//`.\n- Keep `SKILL.md` concise; move detailed workflows to `references/`.\n- Put reusable test paths in `cases/*.yaml`.\n- New or edited cases must include `priority`, `risk`, `ci_eligible`, and `evidence_required` so agents can select the right test set without rereading every file.\n- Use `env_any` / `automation_env_any` for one-of machine inputs, such as `LANGBOT_PIPELINE_URL|LANGBOT_PIPELINE_NAME`; do not list those alternatives as separate all-required env keys.\n- Put reusable groups of cases in `suites/*.yaml` rather than hardcoding test sets in docs or CLI code.\n- Put growing failure knowledge in `troubleshooting/*.yaml`.\n- Do not hardcode local ports in testing docs; use `skills/.env` variables and machine-local `skills/.env.local` overrides.\n- Do not store secrets, API keys, OAuth tokens, or localStorage token values.\n\n## Required Checks\n\nAfter structural changes, run:\n\n```bash\nbin/lbs validate\n```\n\nAfter changing skills, cases, or troubleshooting assets, run:\n\n```bash\nbin/lbs index\n```\n\nUse `bin/lbs env show` to inspect defaults and `bin/lbs env doctor` when diagnosing local environment readiness. Env output is redacted by default; do not work around that by printing raw secrets.\n`bin/lbs` is a generated local wrapper. If it is missing on a fresh checkout, run `npm run bootstrap` from this directory first; `npm install` also regenerates it via `prepare`.\nUse `bin/lbs fixture check` before fixture-heavy cases such as MCP, RAG, multimodal, or plugin smoke tests.\nUse `bin/lbs case list --ready` for cases that have no missing machine inputs and no manual preconditions. Use `bin/lbs case list --machine-ready` when you want to keep `manual-check` candidates and confirm their preconditions yourself.\n\nBefore executing a saved QA path, generate the agent-facing plan:\n\n```bash\nbin/lbs test plan \n```\n\nRead the plan readiness sections before running the browser path. Missing env,\nautomation env, or fixture readiness means the case is not ready to execute and\nshould be marked `blocked` or fixed first.\n`manual_check` means machine inputs are present but the agent must verify the\ndeclared `preconditions` or `setup` items before executing the UI path. Do not\nturn a `manual_check` case into `pass` until those items were checked in the\nsame run.\n\nBefore executing a group of saved QA paths, generate the suite plan:\n\n```bash\nbin/lbs suite plan \n```\n\nUse `bin/lbs suite start ` to create a shared suite run id, suite evidence root, per-case evidence directories, and `suite-start.json`/`suite-start.md` handoff files. Then run `bin/lbs suite report --evidence-dir ` to aggregate case results.\nAutomation scripts write `automation-result.json`; write the final per-case `result.json` with `bin/lbs test result --result --reason --evidence-dir --evidence ` after collecting the required evidence. A `pass` result must include all required evidence.\nFor runner-specific Debug Chat cases, prefer case-specific pipeline env keys such as `LANGBOT_LOCAL_AGENT_PIPELINE_URL` over the generic `LANGBOT_PIPELINE_URL`; otherwise an agent can accidentally test the wrong runner.\n\n\nAGENTS.md\n\n\n# LangBot QA Skills User Guide\n\nUse this guide as the first operational path after reading `README.md` and\n`AGENTS.md`.\n\n## 1. Configure Local Inputs\n\nRead `skills/.env`, then create `skills/.env.local` for machine-local values.\nDo not commit `.env.local`, browser profiles, reports, tokens, API keys, OAuth\nstate, or provider credentials.\n\nMinimum local fields for live browser QA:\n\n```bash\nLANGBOT_REPO=/path/to/LangBot\nLANGBOT_WEB_REPO=/path/to/LangBot/web\nLANGBOT_BACKEND_URL=http://127.0.0.1:5300\nLANGBOT_FRONTEND_URL=http://127.0.0.1:3000\nLANGBOT_DEV_FRONTEND_URL=http://127.0.0.1:3000\nLANGBOT_BROWSER_PROFILE=/path/to/langbot-browser-profile\nLANGBOT_CHROMIUM_EXECUTABLE=/path/to/chromium-or-playwright-chrome\nLANGBOT_E2E_LOGIN_USER=qa-local@example.com\n```\n\n`LANGBOT_E2E_LOGIN_USER` is a local QA account. The setup automation uses the\nLangBot recovery key from the active checkout to initialize or refresh that\nlocal account and write a browser `localStorage` token. It does not need the\nuser's GitHub or Space credentials.\n\n## 2. Check Readiness\n\nFrom `skills/`:\n\n```bash\nbin/lbs env show\nbin/lbs env doctor\nbin/lbs validate\nbin/lbs index --check\n```\n\n`env doctor` should report reachable backend and frontend URLs before live\nbrowser cases are run. Missing Space provider credentials are not a LangBot\nproduct pass; classify them as `env_issue` and configure the local Space\nprovider before measuring Debug Chat performance.\n\n## 3. Start Services\n\nStart the backend from `LANGBOT_REPO`:\n\n```bash\ncd \"$LANGBOT_REPO\"\nuv run main.py\n```\n\nStart the standalone frontend from `LANGBOT_WEB_REPO` and point it at the\nbackend:\n\n```bash\ncd \"$LANGBOT_WEB_REPO\"\nVITE_API_BASE_URL=\"$LANGBOT_BACKEND_URL\" pnpm dev --host 0.0.0.0\n```\n\nIf `VITE_API_BASE_URL` is missing, browser tests can load the Vite page but send\nAPI requests to the frontend port, which produces false UI failures.\n\n## 4. Prepare User-Path Fixtures\n\nFor local-agent Debug Chat cases and the user-path performance gate:\n\n```bash\nnode scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\n```\n\nThe script:\n\n- refreshes the local QA login and browser token;\n- marks the local wizard as skipped;\n- creates or updates a local QA pipeline;\n- scans Space LLM models, tests candidates, and switches to the first working\n Space model with tested fallback models;\n- writes `LANGBOT_PIPELINE_URL`, `LANGBOT_PIPELINE_NAME`, and local-agent\n pipeline/model variables into `skills/.env.local`;\n- returns `env_issue` when no Space model can be scanned or tested.\n\nUseful model controls:\n\n```bash\nLANGBOT_E2E_MODEL_TEST_LIMIT=8\nLANGBOT_E2E_MODEL_FALLBACK_COUNT=3\nLANGBOT_E2E_SKIP_MODEL_UUIDS=uuid-a,uuid-b\nLANGBOT_E2E_SKIP_MODEL_NAMES=model-a,model-b\nLANGBOT_E2E_SCAN_SPACE_MODELS=true\n```\n\nThe setup writes a current-runtime compatibility `max-round` value into the\npipeline config because this backend still reads that field directly during\nmessage truncation. Do not treat it as a long-term QA contract.\n\n## 5. Run Gates\n\nCross-repository contract gate (no model calls):\n\n```bash\nbin/lbs suite plan langbot-workspace-contract-gate\nbin/lbs suite run langbot-workspace-contract-gate --run-id langbot-workspace-contract-local\n```\n\nTop-down workspace release gate (browser workflows plus one complex Agent task):\n\n```bash\nbin/lbs suite plan langbot-workspace-release-gate\nbin/lbs suite run langbot-workspace-release-gate --run-id langbot-workspace-release-local --include-manual-check\n```\n\nRun `agent-run-ledger-audit` immediately after the complex Agent case. Set\n`LANGBOT_AGENT_RUN_ID` when the test instance has concurrent operator traffic.\n\nFast contract gate, no live service required:\n\n```bash\nbin/lbs suite run langbot-performance-contract-gate --run-id langbot-contract-local\n```\n\nLive backend gate:\n\n```bash\nbin/lbs suite run langbot-live-backend-gate --run-id langbot-backend-local\n```\n\nBrowser-visible user-path performance gate:\n\n```bash\nbin/lbs suite plan langbot-user-path-performance-gate\nbin/lbs suite run langbot-user-path-performance-gate --run-id langbot-user-path-local --include-manual-check\n```\n\nControlled Debug Chat message-path load gate (manual/non-required; run fake-provider cases serially when they share `LANGBOT_FAKE_PROVIDER_URL`):\n\n```bash\nbin/lbs suite plan langbot-debug-chat-load-gate\nbin/lbs test run langbot-fake-provider-debug-chat-load --run-id langbot-fake-load-local\nbin/lbs test run langbot-fake-provider-debug-chat-slow-load --run-id langbot-fake-slow-local\nbin/lbs test run langbot-fake-provider-debug-chat-fault-recovery --run-id langbot-fake-fault-local\nbin/lbs test run langbot-space-debug-chat-concurrency-smoke --run-id langbot-space-smoke-local\n```\n\nCross-pipeline Debug Chat isolation is a separate manual regression gate because\ncurrent releases may fail it due to product bug #2286:\n\n```bash\nbin/lbs suite plan langbot-debug-chat-isolation-gate\nbin/lbs suite run langbot-debug-chat-isolation-gate --run-id langbot-debug-chat-isolation-local --include-manual-check\n```\n\nStart with `langbot-fake-provider-debug-chat-load`. It launches a local\nOpenAI-compatible fake provider, creates the matching provider/model/pipeline,\nthen sends concurrent WebSocket Debug Chat messages through the real backend.\nUse `langbot-fake-provider-debug-chat-slow-load` to measure the same path under\ndeterministic streaming latency. Use\n`langbot-fake-provider-debug-chat-fault-recovery` to inject bounded provider\nHTTP failures and confirm later Debug Chat requests recover. Use the separate\n`langbot-debug-chat-isolation-gate` to verify that concurrent Debug Chat traffic\non two pipelines does not leak assistant responses across pipeline boundaries;\ncurrent releases may fail that gate because of #2286, so keep it out of the\nnormal load gate until the product fix lands.\nUse `langbot-space-debug-chat-concurrency-smoke` only as a low-volume live\nprovider smoke; it includes Space/model/network latency and should be compared\nagainst the fake-provider baseline before attributing failures to LangBot.\n\n`manual_check` means the agent must confirm the declared preconditions for that\nrun window. When setup automation is declared, run output may stop early with\n`env_issue`; fix that environment input before treating the product path as\nmeasured.\n\n## 6. Read Results\n\nSuite reports live under `skills/reports/`. Evidence lives under\n`skills/reports/evidence//`.\n\nFor performance cases, inspect:\n\n- `metrics.json` for p50/p95/p99, error rate, and total duration;\n- `automation-result.json` for threshold decisions and artifacts;\n- `console.log` and `network.log` for frontend/API failures;\n- backend logs for provider, runner, WebSocket, or persistence failures.\n\nDo not call a user-path performance result a LangBot overhead regression until\nprovider/tool/network time has been separated or ruled out.\n\n\n{\n \"private\": true,\n \"type\": \"module\",\n \"bin\": {\n \"lbs\": \"./bin/lbs\"\n },\n \"scripts\": {\n \"bootstrap\": \"node scripts/bootstrap-lbs.mjs\",\n \"prepare\": \"node scripts/bootstrap-lbs.mjs\",\n \"prevalidate\": \"node scripts/bootstrap-lbs.mjs\",\n \"preindex\": \"node scripts/bootstrap-lbs.mjs\",\n \"preindex:check\": \"node scripts/bootstrap-lbs.mjs\",\n \"pretest\": \"node scripts/bootstrap-lbs.mjs\",\n \"precheck\": \"node scripts/bootstrap-lbs.mjs\",\n \"lbs\": \"node src/lbs.ts\",\n \"test\": \"node test/lbs-cli.test.ts && python3 -m unittest discover -s test -p 'test_*.py'\",\n \"validate\": \"node src/lbs.ts validate\",\n \"index\": \"node src/lbs.ts index\",\n \"index:check\": \"node src/lbs.ts index --check\",\n \"check:syntax\": \"find src test scripts -type f \\\\( -name '*.ts' -o -name '*.mjs' \\\\) -print0 | xargs -0 -n1 node --check\",\n \"check\": \"npm run check:syntax && npm run validate && npm test\"\n },\n \"engines\": {\n \"node\": \">=22.6\"\n },\n \"devDependencies\": {\n \"playwright\": \"^1.60.0\"\n }\n}\n\n\n# LangBot Agent Testing 技术选型\n\n## 状态\n\n这是技术选型背景文档，不是当前路线图。当前黑盒 E2E QA 的实施顺序见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n## 目标\n\n`langbot-skills` 的目标不是替代测试框架，而是沉淀 agent 可复用的测试资产，让开发者 clone 仓库后，可以让 Codex、Claude Code、Computer Use 或 Playwright MCP 复用已有路径完成 LangBot 功能验证。\n\n核心原则：\n\n- Skill 负责路由和少量规则。\n- Reference 负责可读流程和背景知识。\n- Case 负责结构化测试路径。\n- Troubleshooting 负责结构化故障资产。\n- `lbs` 负责结构校验、索引、资产创建和未来的运行/报告能力。\n- UI/browser 是产品 QA 的主路径；API/curl 只用于诊断。\n\n## 浏览器控制层\n\n不同开发者可用的浏览器控制能力不同，所以浏览器层必须可替换。\n\n| 方案 | 适用场景 | 优点 | 代价 |\n|---|---|---|---|\n| Codex / Claude Computer Use | agent 可以直接控制可见浏览器 | 登录和交互路径最自然，通常不需要额外 MCP 浏览器桥接 | 依赖具体 agent 工具能力 |\n| Playwright MCP | 没有 Computer Use，但有 MCP 浏览器工具 | 稳定、可脚本化、适合回归路径 | OAuth 登录通常需要额外 visible profile |\n| 直接 Playwright 脚本 | 测试路径非常稳定，适合 CI | 可重复性强 | 需要维护脚本和 selector |\n| 商业 AI QA 平台 | 团队希望外包测试运行平台 | 报告和 PR 集成完整 | 成本和平台绑定 |\n\n## 当前推荐\n\n先采用分层降级：\n\n```text\n有 Computer Use？\n 是 -> 使用 Computer Use 控制浏览器\n 否 -> 使用 Playwright MCP\n\n需要 GitHub OAuth？\n 是 -> 使用持久浏览器 profile，让用户手动完成登录\n 否 -> 直接使用已有登录态或测试账号状态\n```\n\n具体选择逻辑沉淀在：\n\n```text\nskills/langbot-env-setup/references/browser-access-selection.md\n```\n\n测试原则固定在：\n\n```text\ndocs/qa-agent/03-agent-browser-qa-principles.md\n```\n\n## 环境变量层\n\n测试文档不应写死端口。共享默认值放在：\n\n```text\nskills/.env\n```\n\n关键变量：\n\n```text\nLANGBOT_FRONTEND_URL\nLANGBOT_BACKEND_URL\nLANGBOT_DEV_FRONTEND_URL\nLANGBOT_REPO\nLANGBOT_WEB_REPO\nLANGBOT_BROWSER_PROFILE\n```\n\nAgent 执行测试前应先读取 `skills/.env`，再用用户提供的当前环境或已启动服务覆盖默认值。\n\n## 测试资产层\n\n测试资产分两类：\n\n```text\nskills//\n references/ # Markdown 流程说明\n cases/ # 结构化测试用例\n troubleshooting/ # 结构化故障记录\n```\n\n当前已实现：\n\n- `SKILL.md` 路由\n- `references/*.md`\n- `lbs case new/list/show`\n- `lbs trouble show/search`\n- `lbs test plan`\n- `lbs test report`\n- `lbs list / validate / index`\n\n下一步重点：\n\n- 日志守卫规则补充\n- 报告产物管理\n\n## 关键判断\n\n不要强制所有内容只能通过 CLI 修改。更好的模式是：\n\n- 新增 case/troubleshooting：优先使用 `lbs`\n- 大段流程说明：允许直接编辑 Markdown\n- 结构性变更后：必须运行 `lbs validate`\n- 任何生成索引的变更后：运行 `lbs index`\n\n这样既能沉淀结构化资产，又不会在 schema 未稳定时拖慢迭代。\n\n\n# LangBot Skills 测试资产库规划\n\n## 状态\n\n这是早期测试资产库规划文档，保留用于解释 `langbot-skills` 的分层来源。\n\n当前路线已经收敛为黑盒 E2E QA：开发者用 agent 通过浏览器测试 LangBot，\n稳定路径沉淀为 case，失败知识沉淀为 troubleshooting。`lbs test report` 和\n日志守卫已有 MVP，后续重点是报告证据、case 元数据和少量稳定路径自动化。当前优先级见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n本文中关于 `case list/show`、`trouble show/search`、`test plan` 的“计划实现”\n内容已经部分过时，因为这些能力已经落地。\n\n## 目标\n\n让开发者 clone `langbot-skills` 后，可以把测试意图交给 agent，由 agent 复用已有环境配置、测试路径和故障知识完成 LangBot 功能验证。\n\n典型场景：\n\n- 冒烟测试：验证 pipeline Debug Chat、provider、常见页面是否正常。\n- Provider 测试：添加 DeepSeek/OpenAI/Claude 等供应商并验证模型可用。\n- 新 feature 测试：探索新 UI 路径，并在稳定后沉淀成 case/reference。\n- 回归测试：复用旧路径，避免每个窗口重新探索登录、模型配置、pipeline 调试。\n- 故障沉淀：把 runtime 超时、代理不一致、WebSocket 问题记录为可搜索资产。\n\n核心方向见 `03-agent-browser-qa-principles.md`：agent 必须以浏览器/UI 为主路径，API/curl 只能作为诊断手段。\n\n## 当前仓库结构\n\n```text\nskills/\n .env # 共享默认变量\n langbot-env-setup/ # 环境准备、浏览器控制路径、代理、登录态\n langbot-testing/ # WebUI / provider / pipeline 测试入口\n langbot-plugin-dev/ # 插件开发测试\n langbot-eba-adapter-dev/ # 平台适配器开发测试\nsrc/\n lbs.ts # CLI 源码\nbin/\n lbs # CLI 入口\ndocs/\n qa-agent/ # 规划文档，历史目录名保留\n```\n\n## 设计分层\n\n### 1. Skill 层\n\n`SKILL.md` 只做触发和路由，不承载大段流程。\n\n例子：\n\n```text\nlangbot-env-setup -> 选择 Computer Use / Playwright MCP / OAuth profile / proxy\nlangbot-testing -> 选择 WebUI / pipeline / provider / troubleshooting\n```\n\n### 2. Reference 层\n\nMarkdown 记录人和 agent 都能读的流程说明。\n\n适合内容：\n\n- 如何选择浏览器控制方式\n- 如何启动/检查服务\n- 如何执行 pipeline Debug Chat\n- 如何处理 OAuth 登录态\n\n### 3. Case 层\n\n使用 YAML 记录可重复测试路径。\n\n建议结构：\n\n```text\nskills/langbot-testing/cases/\n pipeline-debug-chat.yaml\n provider-deepseek.yaml\n```\n\n建议格式：\n\n```yaml\nid: pipeline-debug-chat\ntitle: Pipeline Debug Chat returns a bot response\nmode: agent-browser\narea: pipeline\ntype: smoke\nskills:\n - langbot-env-setup\n - langbot-testing\nenv:\n - LANGBOT_FRONTEND_URL\n - LANGBOT_BACKEND_URL\nsteps:\n - Open LANGBOT_FRONTEND_URL\n - Navigate to Pipelines\n - Open target pipeline\n - Select Debug Chat\n - Send deterministic prompt\nchecks:\n - \"UI: User message appears\"\n - \"UI: Bot message appears\"\n - \"Console: No unexpected frontend errors\"\n - \"Logs: Backend log includes Conversation(0) Streaming completed\"\ndiagnostics:\n - \"Use API/curl only after the UI path is attempted, to distinguish frontend display failure from backend/runtime failure.\"\ntroubleshooting:\n - plugin-runtime-timeout\n - proxy-env-mismatch\n```\n\n### 4. Troubleshooting 层\n\n故障资产会逐渐变大，适合结构化记录。\n\n历史 Markdown 入口保留在：\n\n```text\nskills/langbot-testing/references/troubleshooting.md\n```\n\n当前 canonical 结构化故障资产在：\n\n```text\nskills/langbot-testing/troubleshooting/\n plugin-runtime-timeout.yaml\n proxy-env-mismatch.yaml\n```\n\n### 5. CLI 层\n\n`lbs` 是统一入口，不再引入独立 `qa` 命令。\n\n已实现或当前可用：\n\n```bash\nbin/lbs list\nbin/lbs validate\nbin/lbs index\nbin/lbs new-skill \nbin/lbs new-ref \nbin/lbs case new pipeline-debug-chat --title \"Pipeline Debug Chat\"\nbin/lbs case list\nbin/lbs case show pipeline-debug-chat\nbin/lbs trouble list \nbin/lbs trouble show plugin-runtime-timeout\nbin/lbs trouble search runtime\nbin/lbs trouble add --title ... --symptom ... --cause ... --fix ...\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log\n```\n\n## 测试库位置\n\n不要使用隐藏 `.qa/` 作为主测试库。测试资产应该和 skill 放在一起，便于触发和维护：\n\n```text\nskills/langbot-testing/\n references/\n cases/\n troubleshooting/\n reports/ # 可选，本地运行产物可按需忽略或输出到外部目录\n```\n\n如果未来需要项目本地测试库，可以允许 `lbs` 支持 `--workspace` 或项目根目录配置，但 canonical 资产仍保存在 `langbot-skills`。\n\n## 阶段规划\n\n### 阶段一：环境和测试路径沉淀\n\n状态：基本完成，持续维护。\n\n- `skills/.env` 管共享默认变量。\n- `langbot-env-setup` 拆出 Computer Use、Playwright MCP、OAuth profile、proxy、service startup。\n- `langbot-testing` 记录 WebUI、pipeline、provider 测试路径。\n- `lbs validate/index` 维护结构。\n\n完成标准：\n\n- agent 可以从 `skills/.env` 和 references 中找到当前测试入口。\n- pipeline Debug Chat 这类路径不再需要从头探索。\n\n### 阶段二：结构化 case/troubleshooting\n\n状态：主体已完成，继续补齐元数据和资产质量。\n\n目标：\n\n- `lbs case new/list/show`\n- `lbs trouble show/search`\n- case id 去重、字段校验、索引生成\n\n完成标准：\n\n- 冒烟测试路径可以用结构化 case 表示。\n- 下一个 agent 窗口可以直接读取 case 执行。\n\n### 阶段三：计划和报告\n\n状态：已有 MVP，继续完善。\n\n目标：\n\n- `lbs test plan `\n- agent 按 plan 使用浏览器执行 UI QA\n- `lbs test report`\n- 日志守卫集成\n- 报告产物和 evidence 约定\n\n完成标准：\n\n- agent 可以按 case plan 执行浏览器测试。\n- 结果报告包含 UI 结果、后端日志、console 错误和 troubleshooting 建议。\n\n## 执行规则\n\n- agent 可以直接编辑 Markdown reference。\n- 新增结构化 case/troubleshooting 时，优先使用 `lbs`。\n- 每次结构变更后运行 `bin/lbs validate`。\n- 每次索引相关变更后运行 `bin/lbs index`。\n- 测试文档不写死端口，使用 `skills/.env` 中的 URL 变量。\n- 测试 case 的 `mode` 固定为 `agent-browser`。\n- API/curl 只能写入 `diagnostics`，不能替代 UI 步骤和 UI 检查。\n\n\n# 日志守卫规划\n\n## 状态\n\n这是当前活跃设计，已有第一版文件扫描 MVP。实现边界需要和黑盒 E2E 路线保持一致：\n\n- 日志守卫服务于 `lbs test report`。\n- 它不替代浏览器/UI 判断。\n- 它不发展成独立后端 API 测试框架。\n- 第一版默认扫描 `LANGBOT_REPO/data/logs/` 下最新的 `langbot-*.log`，也可扫描 agent\n 显式提供的 backend/frontend/console 日志文件。\n\n当前总体路线见：\n\n```text\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n## 目标\n\n日志守卫是 `lbs test report` 的一部分，用来在 agent 执行测试期间捕获 UI 断言之外的运行时问题。\n\n当前命令方向已收敛为 `lbs test plan` / `lbs test report`。日志守卫服务于 agent-browser QA，不是独立的后端 API 测试入口。\n\nLangBot 是异步且集成度高的系统，有些问题不会直接表现为页面失败：\n\n- 后台任务异常\n- 未等待的协程\n- Provider 流式调用失败\n- 插件 runtime 超时\n- 平台发送失败\n- 数据库连接问题\n- 敏感信息泄露\n\n日志守卫负责把这些信号结构化地放进测试报告，并关联到 troubleshooting 资产。\n\n## 输入\n\n日志守卫应从环境和运行上下文读取配置：\n\n- `skills/.env` 中的 `LANGBOT_BACKEND_URL`\n- `skills/.env` 中的 `LANGBOT_REPO`，用于自动发现 LangBot 后端日志\n- `lbs test plan` / report 记录的 case id\n- LangBot 后端进程输出\n- 前端 dev server 输出\n- 浏览器 console/network 错误\n- case 声明的 success/failure patterns 和 expected failures\n\n## MVP 范围\n\n- 读取一个或多个日志流或日志文件。\n- 检测错误模式。\n- 支持按 case id 或 pattern 白名单。\n- 输出 JSON/Markdown 摘要。\n- 发现非预期错误时让测试报告标记失败；未来如果有自动执行器，再返回非零退出码。\n\n## 错误分类\n\n### 永远非预期\n\n除非 case 明确声明，否则应失败：\n\n- `Traceback`\n- `Task exception was never retrieved`\n- `RuntimeWarning: coroutine .* was never awaited`\n- `Unclosed client session`\n- `Unclosed connector`\n- `KeyError`\n- `TypeError`\n- `AttributeError`\n- 密钥、token、secret 明文泄露\n\n### Case 预期错误\n\n只有当前 case 声明时允许：\n\n- 无效 provider key\n- Provider 认证失败\n- 无效 webhook payload\n- 插件测试故意抛错\n- 超时测试\n- 限流测试\n\n### 仅警告\n\n报告但默认不失败：\n\n- 可恢复重试\n- 恢复的超时\n- 废弃配置\n- 慢请求\n- 版本检查失败\n\n## 与 Troubleshooting 集成\n\n日志守卫不只输出错误文本，还应尽量匹配已知 troubleshooting id。\n\n例子：\n\n```text\nAction list_plugins call timed out\nAction list_runners call timed out\nAction invoke_llm_stream call timed out\n```\n\n可映射到：\n\n```text\nplugin-runtime-timeout\n```\n\n```text\nuppercase proxy points to one host, lowercase proxy points to another\n```\n\n可映射到：\n\n```text\nproxy-env-mismatch\n```\n\n## 未来命令\n\n```bash\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report --output report.md\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log --console-log /path/to/console.log\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs test report pipeline-debug-chat --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\" --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --no-auto-log\n```\n\n运行报告应包含：\n\n- case id\n- URL 和环境变量摘要，不能包含 secrets\n- 浏览器可见结果\n- 后端日志摘要\n- console/network 错误\n- 匹配到的 troubleshooting id\n- 通过/失败结论\n\n## MVP 完成标准\n\n- 可以自动扫描最新 LangBot 后端日志，也可以扫描前端日志和 console 日志文件。\n- 可以用 `--since` 或 `--tail-lines` 把扫描范围限制到本次测试窗口。\n- 可以检测明显 Python/运行时错误和 secret 泄露风险。\n- 可以识别 case 声明的 success/failure patterns。\n- 可以识别 troubleshooting pattern，包括 `plugin-runtime-timeout` 和 `proxy-env-mismatch`。\n- 支持 case 级白名单。\n- 输出机器可读摘要。\n- 至少一个 `langbot-testing` case 使用它。\n\n当前 MVP 已覆盖自动发现 LangBot 后端日志、文件扫描、`--since`/`--tail-lines` 扫描窗口、\n基础错误检测、case success/failure signal、troubleshooting 匹配、secret 脱敏和 `--json`\n输出。仍待继续完善的是 live log 采集、更多规则、case 级 expected failure 的资产化和真实\nE2E report 样例。\n\n\n# Agent Browser QA Principles\n\nThis document fixes the direction of LangBot agent testing so the project does not drift into a backend API smoke-test framework.\n\n## Primary Goal\n\n`langbot-skills` should help an agent behave like a QA engineer using the product, not like a backend curl script.\n\nThe primary path is:\n\n```text\ndeveloper intent -> lbs test plan -> agent controls browser -> UI result + console + logs -> report/assets\n```\n\n## Rules\n\n1. Browser/UI interaction is the source of truth for product QA cases.\n2. A backend API or curl response is never enough to mark a UI case passed.\n3. API/curl/log checks are allowed as diagnostics after a UI path is attempted or when debugging environment readiness.\n4. A case passes only when the user-visible UI result is correct.\n5. The agent should inspect browser console/network output when available.\n6. If screenshot or vision capability is available, the agent should check for blank pages, overlap, hidden actions, broken layout, and error toasts.\n7. If no visual model is available, use DOM/accessibility snapshots and console output instead.\n8. New stable UI paths should be added as `cases/*.yaml`.\n9. New recurring failure modes should be added as `troubleshooting/*.yaml`.\n10. Secrets, tokens, API keys, and localStorage token values must never be printed.\n\n## Command Semantics\n\n`lbs` manages assets and produces plans. It does not replace the agent's browser-control ability.\n\n```bash\nbin/lbs test plan pipeline-debug-chat\n```\n\nThis command outputs:\n\n- environment variables to use\n- required skills\n- browser steps\n- UI/console/visual/log checks\n- diagnostic options\n- related troubleshooting patterns\n- report template\n\nThe active agent then executes the plan with Computer Use, Playwright MCP, or another available browser-control tool.\n\n## Diagnostics\n\nDiagnostics can include:\n\n- `bin/lbs env doctor`\n- browser console/network inspection\n- backend logs\n- targeted API/curl checks\n\nDiagnostics answer \"where did it fail?\" They do not replace \"did the user-visible UI work?\"\n\n\n# 黑盒 E2E QA 路线图\n\n## 定位\n\nLangBot 有大量外部依赖：模型供应商、plugin runtime、浏览器登录态、\nmarketplace/network、RAG engine、sandbox backend、平台适配器等。单测仍然有价值，\n但这个 QA 方向当前不优先解决 LangBot core 的单测覆盖率问题，因为重 mock 往往不能\n真实代表产品路径。\n\n`langbot-skills` 当前目标是让黑盒 E2E 测试变得可执行、可沉淀、可复用：\n\n```text\n开发者测试意图\n-> 复用或新增 case\n-> agent 通过浏览器执行\n-> UI + console + network + log 证据\n-> report\n-> 反哺 case / troubleshooting\n```\n\n这是面向开发者的 QA 资产库。开发者可以让 agent 测一个 feature；如果路径稳定，\n就把路径正规化为 case，让下一个开发者或 QA agent 继续复用。\n\n## 非目标\n\n- 这一阶段不优先建设 LangBot core 单测覆盖率。\n- 不把 API/curl 作为 WebUI 行为的通过标准。\n- 不要求每个 case 都能进 CI。\n- 不在 report 和日志守卫有用之前急着做完整 browser runner。\n- 不把外部 provider、OAuth、marketplace 抖动直接判成产品失败，除非证据明确。\n\n## 当前状态\n\n仓库已经具备第一层基础设施：\n\n- `skills/.env` 和 `skills/.env.local` 管理测试环境；\n- `langbot-env-setup`、`langbot-testing`、`langbot-plugin-dev` 等 skill；\n- `skills/langbot-testing/cases` 下的结构化 case；\n- `skills/langbot-testing/troubleshooting` 下的结构化故障资产；\n- RAG、多模态、plugin、MCP 等 fixture；\n- `bin/lbs validate`、`bin/lbs index`、`bin/lbs case`、`bin/lbs trouble`、\n `bin/lbs test plan`、`bin/lbs test start`、`bin/lbs test report`。\n\n所以当前已经不是“先把路径写进 Markdown”的阶段，而是进入“让每次运行有证据、\n有报告、能沉淀”的阶段。\n\n## 测试模型\n\nUI case 只有在用户可见行为正确时才能通过。辅助证据必须解释同一次运行。\n\n通过一个 UI case 的最低证据：\n\n- 用户可见的成功信号，例如 bot 回复、provider 保存成功、文件上传完成、plugin 页面渲染；\n- 没有意外 browser console error；\n- 相关时间窗口内没有意外后端/runtime 错误；\n- 有截图、DOM snapshot 或同等视觉/结构证据，如果当前 agent 能获取；\n- API/curl 只在解释同一条 UI 路径时作为诊断证据。\n\n失败报告需要保留足够信息，让开发者能复现或分流：\n\n- case id 和实际测试 URL；\n- 使用的 browser path；\n- 最后可见 UI 状态；\n- console/network 症状；\n- 相关后端/前端日志；\n- 匹配到的 troubleshooting id；\n- 这是产品失败、环境问题、外部依赖抖动，还是证据不足。\n\n## 结果词汇\n\n统一使用这些结果：\n\n- `pass`：UI 行为正确，辅助证据干净。\n- `fail`：UI 行为错误，或同一次运行的 console/log 出现意外产品错误。\n- `blocked`：缺登录、缺 provider credentials、服务未启动等原因导致目标路径没有跑起来。\n- `env_issue`：失败在目标行为之外，例如 proxy、OAuth、provider quota、marketplace outage、\n 本地服务启动问题。\n- `flaky`：同一环境下结果不稳定，进入门禁前需要先稳定。\n\n做 merge/release 判断时，`env_issue` 和 `blocked` 不能算产品通过。\n\n## 路线图\n\n### Phase 0：对齐文档\n\n目标：明确当前黑盒 E2E 方向。\n\n交付物：\n\n- `docs/qa-agent/README.md` 文档状态导航；\n- 本路线图；\n- 给旧规划文档加状态说明。\n\n完成标准：\n\n- 新贡献者不用通读所有旧文档，也能知道当前重点。\n\n### Phase 1：Test Report MVP\n\n状态：已有第一版。\n\n目标：让每次 agent browser 测试都有一致报告格式，即使 browser 执行还没自动化。\n\n建议命令：\n\n```bash\nbin/lbs test start \nbin/lbs test report --output reports/-.md\n```\n\nMVP 行为：\n\n- 读取 case 和关联 troubleshooting；\n- 生成 Markdown report 模板；\n- 生成 run handoff，固定本次测试的 start timestamp 和推荐 report command；\n- 写入脱敏后的环境摘要；\n- 提供 `pass/fail/blocked/env_issue/flaky` 结果选项；\n- 包含 UI result、console errors、network symptoms、logs、screenshots、\n diagnostics、matched troubleshooting、assets to update 等 section；\n- 支持 `--json`，输出机器可读报告。\n\n第一版已经是 report generator，不急着做自动判定。先把 evidence 收集格式统一起来，\n再做自动化更稳。\n\n完成标准：\n\n- agent 可以先跑 `lbs test start `，用它给出的时间窗口执行浏览器路径，\n 然后按固定格式填写 report，不需要每次重新发明报告结构。\n\n### Phase 2：日志守卫 MVP\n\n状态：已有第一版文件扫描。\n\n目标：捕获 UI 不一定明显展示的 runtime 问题。\n\n日志守卫应集成进 `lbs test report`，不要发展成独立后端 API 测试框架。\n\n建议命令形态：\n\n```bash\nbin/lbs test report \\\n --backend-log /path/to/backend.log \\\n --frontend-log /path/to/frontend.log \\\n --console-log /path/to/console.log \\\n --evidence-dir reports/evidence/ \\\n --since \"2026-05-21T10:30:00+08:00\" \\\n --tail-lines 2000 \\\n --output reports/-.md\n```\n\nMVP 行为：\n\n- 默认从 `LANGBOT_REPO/data/logs/` 扫描最新 `langbot-*.log`；\n- 支持 agent 显式提供 backend、frontend、console 日志文件；\n- 支持读取 evidence 目录下的 `automation-result.json`，把浏览器自动化脚本结论纳入报告；\n- 支持 `lbs test result` 为人工/agent browser 运行写入标准 `result.json`，供 suite 聚合；\n- 支持 `--since` 和 `--tail-lines`，避免历史日志污染本次报告；\n- 检测默认非预期模式，例如 `Traceback`、未 await coroutine、unclosed client/connector、\n `KeyError`、`TypeError`、`AttributeError`、明显 secret 泄露；\n- 匹配 case 声明的 `success_patterns` 和 `failure_patterns`；\n- 匹配已知 troubleshooting，先支持 `plugin-runtime-timeout` 和 `proxy-env-mismatch`；\n- 只有 case 明确声明时，才允许 expected failure；\n- 将发现分类为 fail、warning、matched troubleshooting、ignored expected issue；\n- 永远不打印 secret 值。\n\n完成标准：\n\n- 至少 `pipeline-debug-chat` 能生成包含日志摘要和 troubleshooting 匹配结果的 report。\n\n### Phase 3：Case 元数据加固\n\n状态：已有第一版。\n\n目标：让 case 更容易选择、执行和晋级。\n\n字段逐步补充，保持向后兼容：\n\n```yaml\npriority: p0 | p1 | p2\nrisk: low | medium | high\nci_eligible: false\npreconditions:\n - \"Authenticated browser profile is available.\"\nsetup:\n - \"Start LangBot backend and frontend.\"\ncleanup:\n - \"Remove temporary provider, plugin, or knowledge base if created.\"\nexpected_failures: []\nsuccess_patterns:\n - \"Conversation(0) Streaming completed\"\nfailure_patterns:\n - \"Action invoke_llm_stream call timed out\"\nevidence:\n required:\n - ui\n - console\n - backend_log\n```\n\n当前实现采用扁平字段 `evidence_required`，避免轻量 YAML 解析器在 case 文件里承载嵌套结构。\n`bin/lbs validate` 会校验 `priority`、`risk`、`ci_eligible`、`evidence_required`、\n`automation` 脚本路径、case 关联 skill 和 troubleshooting 交叉引用。`bin/lbs case list`\n支持 `--json`、`--type`、`--area`、`--tag`、`--priority`、`--risk`、`--automation`、`--ci`\n、`--ready` 和 `--machine-ready` 过滤，方便 agent 快速选择测试集。\n`env_any` 和 `automation_env_any` 用于表达 URL-or-name 这类 one-of 输入，避免把可替代变量误判为全部必填。\n\n当前也有 `skills//suites/*.yaml` 和 `bin/lbs suite plan `，用于组织常跑测试集，\n例如 `core-smoke`、`local-agent-gate` 和\n`agent-runner-release-gate`。发布门禁使用 `agent-runner-release-preflight`\n先分类配置 blockers 和 runtime env issues，再运行较重的浏览器 Debug Chat case。\n依赖 fixture 的 case 可以在浏览器执行前先跑 `bin/lbs fixture check`，检查\n`fixtures/fixtures.json` 登记的 deterministic 文件、plugin 包和本地测试 server 是否存在。\n`bin/lbs suite start ` 会生成 suite run id、suite evidence root、per-case evidence 目录、\n`suite-start.json`/`suite-start.md` handoff 文件和 per-case evidence 命令；\n浏览器自动化脚本会写入 `automation-result.json`，供 `bin/lbs test report` 展示原始自动化结论；\n`bin/lbs test result ` 会在人工/agent browser case 完成后写入最终 `result.json`；\n`bin/lbs suite report --evidence-dir ` 会聚合各 case 的 `result.json`，并且\n不会把缺少 required evidence 的 `pass` 当作 suite 通过。\nRunner 专用 Debug Chat case 通过 `automation_pipeline_url_env` 和\n`automation_pipeline_name_env` 绑定专用 pipeline 变量，避免 local-agent、Codex 或\nClaude Code case 误用通用 `LANGBOT_PIPELINE_URL` 后产生假阳性。\nDebug Chat case 还可以通过 `automation_stream_output` 固定流式或非流式发送路径。\n多模态 Debug Chat case 可以通过 `automation_image_base64_fixture` 复用 deterministic 图片 fixture。\n`test plan` 和 `suite plan` 会输出 readiness，让 agent 在执行浏览器前就看到缺失的 env、\n自动化变量、fixture，以及需要人工确认的 `manual_check` 前置条件。\n\n完成标准：\n\n- `lbs case list` 或后续 filter 能回答“smoke 跑哪些”、“哪些适合 CI”、\n “哪些需要真实 provider credentials”。\n\n### Phase 4：开发者沉淀流程\n\n目标：开发者让 agent 测新 feature 后，稳定路径不会丢在聊天记录里。\n\n流程：\n\n1. 开发者要求 agent 通过浏览器测试某个 feature。\n2. agent 先按 UI 主路径探索。\n3. agent 用 `lbs test start` 固定运行窗口，再用 `lbs test report` 写报告。\n4. 如果路径稳定，agent 新增或更新 case。\n5. 如果出现可复用故障，agent 新增或更新 troubleshooting。\n6. agent 跑 `bin/lbs validate` 和 `bin/lbs index`。\n\n完成标准：\n\n- feature QA 的结果能进入资产库，而不是只留在一次对话里。\n\n### Phase 5：选择性浏览器自动化\n\n状态：已有第一版 `test run` 入口和两个 Playwright 脚本。\n\n目标：只自动化少量稳定、值得重复跑的黑盒路径。\n\n建议顺序：\n\n1. `webui-login-state`\n2. `pipeline-debug-chat`\n3. `local-agent-basic-debug-chat`\n4. `local-agent-rag-debug-chat`\n5. 一个基于 deterministic fixture 的 plugin 或 MCP smoke path\n\n执行策略：\n\n- 继续把 Computer Use 或 Playwright MCP 作为默认交互路径；\n- 只给稳定、确定性的路径补直接 Playwright script；\n- 保存 screenshots、console logs、trace/video；\n- flaky 或强依赖真实 credentials 的 provider case 暂时不要进 CI。\n\n当前已经绑定：\n\n- `webui-login-state` -> `scripts/e2e/webui-login-state.mjs`\n- `pipeline-debug-chat` -> `scripts/e2e/pipeline-debug-chat.mjs`\n\n第一版自动化先产出 `reports/evidence//` 下的 console、network、screenshot 和\nresult JSON。真实执行后仍要用 `lbs test report --since ... --console-log ...` 做日志守卫和\n最终报告。开发期间可以先用 `bin/lbs test run --dry-run` 检查命令和 evidence 路径。\nDebug Chat 类脚本应复用 `scripts/e2e/lib/debug-chat.mjs`，避免重复实现 visible response leaf\n判断和已知失败信号分类。\n\n完成标准：\n\n- 小规模 smoke subset 可以不靠人工决定每一步点击；更大的资产库仍然服务于人工/agent\n 驱动的探索式 E2E。\n\n## 下一批动工切片\n\n在做 browser runner 之前，继续做这些：\n\n1. 等 LangBot 当前开发状态稳定后，用一次真实 `pipeline-debug-chat` 跑通\n `test start -> test run -> test report -> test result -> suite report`，产出 sample report。\n2. 只给 smoke/local-agent 首批 case 补必要元数据。\n3. 继续补日志守卫规则，尤其是 WebSocket、plugin runtime、provider streaming、前端\n chunk/rendering failure。\n4. 约定 report 产物目录、截图和 console/network 导出的命名方式。\n5. 再评估是否开始给 `webui-login-state` 和 `pipeline-debug-chat` 做直接 Playwright\n 自动化。\n\n这样 infra 会立刻有用，同时保留后续自动化 browser execution 的空间。\n\n\n# LangBot QA Agent 文档导航\n\n这个目录记录 `langbot-skills` 当前的 QA 方向和后续建设顺序。\n\n## 当前判断\n\n当前重点是 LangBot 的黑盒 E2E QA，不是 LangBot core 的单测覆盖率建设。\n\n`langbot-skills` 要帮助开发者和 QA agent 做接近人工测试的 WebUI 验证：\n\n- 打开真实 LangBot WebUI；\n- 按用户路径点击和输入；\n- 检查用户可见的 UI 结果；\n- 查看 console、network、截图、后端和前端日志；\n- 输出可复用的测试报告；\n- 把稳定 feature 路径沉淀为 case；\n- 把重复故障沉淀为 troubleshooting。\n\nAPI 和 curl 只做诊断。它们可以解释失败原因，但不能让一个 UI case 通过。\n\n## 文档状态\n\n| 文档 | 状态 | 用途 |\n| --- | --- | --- |\n| `04-black-box-e2e-roadmap.md` | 当前主路线图 | 决定下一步建设什么。 |\n| `03-agent-browser-qa-principles.md` | 当前原则文档 | 定义 browser-first QA 的通过标准。 |\n| `02-log-guard-plan.md` | 当前活跃设计 | 设计 `lbs test report` 里的日志守卫。 |\n| `../user-guide.md` | 当前使用手册 | 开发者日常使用。 |\n| `00-technology-options.md` | 背景文档 | 选择 Computer Use、Playwright MCP 或未来直接 Playwright。 |\n| `01-qa-agent-harness-plan.md` | 历史规划，部分过时 | 解释最初分层和目录设计；使用前先看状态说明。 |\n\n## 已过时的点\n\n`01-qa-agent-harness-plan.md` 还保留早期规划状态。现在结构化 cases、\n结构化 troubleshooting、`validate`、`index`、`lbs test plan` 都已经落地。\n\n已经补上第一版 `lbs test start`、`lbs test run`、`lbs test report` 和日志守卫文件扫描。\n`webui-login-state`、`pipeline-debug-chat` 已经绑定直接 Playwright 自动化脚本。后续重点是：\n\n- 报告 evidence 字段继续打磨；\n- case success/failure signal 和日志守卫规则继续补充；\n- 报告产物和 evidence 约定；\n- 等 LangBot 当前开发状态稳定后跑真实 sample report。\n\n不要再把旧阶段列表当成当前 source of truth。后续排序以\n`04-black-box-e2e-roadmap.md` 为准。\n\n\n# LangBot Skills 用户使用手册\n\n## 这个仓库解决什么\n\n`langbot-skills` 是给 agent 使用的 LangBot 测试资产库。开发者 clone 后，可以让 Codex、Claude Code、Computer Use 或 Playwright MCP 复用已有环境配置、测试路径和故障知识，像 QA 一样操作 LangBot WebUI。\n\n核心目标：\n\n- 不让下一个 agent 窗口从头探索登录、模型配置、pipeline 调试。\n- 把稳定 UI 测试路径沉淀为 case。\n- 把常见故障沉淀为 troubleshooting。\n- 让 agent 优先通过浏览器点击验证产品行为。\n- API/curl/log 只作为诊断手段，不作为 UI case 通过标准。\n\n## 快速开始\n\n1. Clone 仓库。\n\n2. 检查本地默认变量：\n\n```bash\nbin/lbs env show\n```\n\n默认变量在：\n\n```text\nskills/.env\n```\n\n本机专用覆盖写到：\n\n```text\nskills/.env.local\n```\n\n它会覆盖 `skills/.env` 中的同名变量，并且不应该提交。\n`skills/.env` 是共享默认值，不应写入本机绝对路径、浏览器 profile、provider key 或其他凭据。\n新机器建议从模板开始：\n\n```bash\ncp skills/.env.example skills/.env.local\n```\n\n常用变量：\n\n```text\nLANGBOT_FRONTEND_URL\nLANGBOT_BACKEND_URL\nLANGBOT_DEV_FRONTEND_URL\nLANGBOT_REPO\nLANGBOT_WEB_REPO\nLANGBOT_BROWSER_PROFILE\n```\n\n3. 检查环境是否就绪：\n\n```bash\nbin/lbs env doctor\nbin/lbs fixture check\n```\n\n`env doctor` 会检查 URL、路径、代理变量等。代理变量是可选项；只有大小写代理变量互相冲突时才会报错。失败不一定代表仓库坏了，通常说明本地 LangBot 没启动、代理不一致或浏览器 profile 不存在。\n`fixture check` 会检查仓库内测试 fixture 是否存在，例如 MCP stdio server、RAG 文档、多模态图片、qa-plugin-smoke 包和 QA Runner 包。它也会校验 `.lbpkg` 是 zip 包，并检查 QA Runner fixture 的入口文件未漂移。\n\n4. 查看已有测试 case：\n\n```bash\nbin/lbs case list\nbin/lbs case list --json --priority p0 --automation\nbin/lbs case list --ready\nbin/lbs case list --machine-ready\nbin/lbs suite list\nbin/lbs suite plan core-smoke\nbin/lbs suite plan agent-runner-release-gate\nbin/lbs suite start core-smoke\nbin/lbs suite start core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\n```\n\n`case list` 支持按 `--type`、`--area`、`--tag`、`--priority`、`--risk`、`--automation`\n、`--ci`、`--ready` 和 `--machine-ready` 过滤。`--ready` 只显示没有缺机器输入且没有人工前置条件的 case；\n`--machine-ready` 过滤掉缺机器输入的 case，但保留 `manual-check`，表示执行前还要确认前置条件。需要交给 agent 自动选择测试集时，优先使用 `--json`，\n其中包含 `priority`、`risk`、`ci_eligible`、`automation`、`evidence_required` 以及\nenv/automation/fixture/manual readiness。\nCase metadata 中的 `env` 和 `automation_env` 表示全部必填；URL 或 name 这类二选一输入会放在\n`env_any` 或 `automation_env_any`，readiness 只要求组合里至少一个变量有值。\n\n如果要跑一组已沉淀的测试路径，优先使用 suite。Suite 位于 `skills//suites/*.yaml`，\n只负责组织 case，不改变 UI/browser 作为通过标准的原则。\n`suite plan` 会聚合 readiness：缺环境变量、缺自动化变量、缺 fixture 或需要\n`manual_check` 时，会在执行前标出受影响的 case。`manual_check` 不是产品通过，\n它表示机器配置已满足但 agent 必须先确认 case 里的 `preconditions` 或 `setup`。\nRunner externalization 发布判断使用 `agent-runner-release-gate`。先跑\n`agent-runner-release-preflight`，把缺 pipeline、runner id 错误、插件未安装这类\n`blocked`，以及 provider key、Box、插件运行时这类 `env_issue` 分开，再执行较重的\n浏览器 Debug Chat case。\n\n5. 生成 agent 执行计划：\n\n```bash\nbin/lbs test plan pipeline-debug-chat\n```\n\n然后把计划交给当前 agent 执行。agent 应使用 Computer Use、Playwright MCP 或其他浏览器控制能力去操作 UI。\n`test plan` 中的 Environment、Automation Readiness、Fixture Readiness 和 Manual\nReadiness 是执行前门禁；如果 readiness 缺失，应先补配置或将本次 case 标记为\n`blocked`。如果状态是 `manual_check`，先确认 `preconditions` 和 `setup`，再开始 UI\n执行。不要把后续 curl/API 诊断当成 UI case 通过。\n\n## 推荐使用方式\n\n### 冒烟测试\n\n你可以直接对 agent 说：\n\n```text\n帮我跑一下 LangBot 新前端 smoke test。\n```\n\nagent 应该：\n\n- 读 `skills/.env`\n- 优先查看 `bin/lbs suite plan core-smoke`，或查找 `type: smoke` 的 cases\n- 生成 test plan\n- 用浏览器执行 UI 操作\n- 检查 console、截图、后端日志\n- 输出简短 QA 报告\n\n### Runner Externalization 发布门禁\n\n你可以直接对 agent 说：\n\n```text\n按 agent-runner release gate 跑完整矩阵，先做 preflight，再跑浏览器 case，并把 blocked/env_issue/fail 分开。\n```\n\nagent 应该先查看 `skills/langbot-testing/references/agent-runner-release-gate.md`，\n再执行：\n\n```bash\nbin/lbs test recommend\nbin/lbs suite plan agent-runner-release-gate\nbin/lbs test run agent-runner-release-preflight --dry-run\nbin/lbs suite start agent-runner-release-gate --run-id agent-runner-release-local --evidence-dir reports/evidence/agent-runner-release-local\n```\n\n`test recommend` 输出的 run 命令默认带 `--dry-run`；确认 readiness 和 `manual_check` 前置条件后，再去掉 `--dry-run` 执行。\n\n完成所有 case 后，用：\n\n```bash\nbin/lbs suite report agent-runner-release-gate --evidence-dir reports/evidence/agent-runner-release-local\n```\n\n没有最终 `result.json`、缺 required evidence、或把 `blocked`/`env_issue` 当成 pass，\n都不能算发布门禁通过。\n\n### 新 Feature 测试\n\n你可以说：\n\n```text\n我改了 provider 页面，帮我测一下 DeepSeek provider 添加、测试、绑定 pipeline 是否正常。\n```\n\nagent 应该：\n\n- 查找相关 case 和 reference\n- 如果没有稳定路径，先探索 UI\n- 用浏览器执行真实交互\n- 失败时用日志/API 辅助定位\n- 稳定后新增或更新 case/reference\n- 新故障沉淀为 troubleshooting\n\n### 定点排错\n\n你可以说：\n\n```text\nDebug Chat 点了没回复，帮我查是前端问题还是后端问题。\n```\n\nagent 应该：\n\n- 先通过 UI 复现\n- 看 console/network\n- 看后端日志\n- 必要时用 API/curl 做诊断\n- 匹配 troubleshooting\n- 给出修复建议或直接修复\n\n## 重要原则\n\n这些原则固定在：\n\n```text\ndocs/qa-agent/03-agent-browser-qa-principles.md\n```\n\n简化版：\n\n- UI/browser 是测试主路径。\n- API/curl/log 只做诊断。\n- 后端接口成功不等于 UI case 通过。\n- case 通过必须以用户可见 UI 结果为准。\n- 有视觉能力时应检查截图。\n- 没有视觉能力时用 DOM/accessibility snapshot 和 console。\n- 不要打印 token、API key、OAuth secret 或 localStorage token 值。\n\n## 规划文档\n\n如果要判断下一步建设什么，先看：\n\n```text\ndocs/qa-agent/README.md\ndocs/qa-agent/04-black-box-e2e-roadmap.md\n```\n\n`01-qa-agent-harness-plan.md` 是早期规划，部分内容已经被当前实现和路线图替代。\n\n## 常用命令\n\n### 环境\n\n```bash\nbin/lbs env show\nbin/lbs env show --json\nbin/lbs env doctor\nbin/lbs fixture list\nbin/lbs fixture check\nbin/lbs fixture check --json\n```\n\n`env show` 和 `env doctor` 默认会对 token、API key、password、secret 以及 URL basic auth\n做脱敏。不要把 `.env.local` 里的原始凭据复制进测试报告。\n\n### Skill 和索引\n\n```bash\nbin/lbs list\nbin/lbs validate\nbin/lbs index --check\nbin/lbs index\n```\n\n### Case\n\n```bash\nbin/lbs case list\nbin/lbs case list --type smoke\nbin/lbs case list --json --priority p1 --tag local-agent\nbin/lbs case list --ready\nbin/lbs case list --machine-ready\nbin/lbs case show pipeline-debug-chat\nbin/lbs case new my-feature --title \"My Feature Works\"\n```\n\n### Suite\n\n```bash\nbin/lbs suite list\nbin/lbs suite list --json --priority p1\nbin/lbs suite show local-agent-gate\nbin/lbs suite plan core-smoke\nbin/lbs suite plan local-agent-gate --json\nbin/lbs suite start core-smoke\nbin/lbs suite start core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\nbin/lbs suite run core-smoke --dry-run --json\nbin/lbs suite run core-smoke --run-id core-smoke-local --evidence-dir reports/evidence/core-smoke-local\nbin/lbs suite start core-smoke --json\nbin/lbs suite report core-smoke --evidence-dir reports/evidence/\nbin/lbs suite report core-smoke --evidence-dir reports/evidence/ --json\nbin/lbs suite new my-feature-gate --title \"My Feature Gate\"\n```\n\n`suite start` 不直接控制浏览器。它生成统一 run id、suite evidence root、每个 case 的 evidence\n目录、`suite-start.json`/`suite-start.md` handoff 文件，以及每个 case 的 `test run`、`test report`\n和 `test result` 命令模板。需要固定路径时，使用 `--run-id` 和 `--evidence-dir`。\n`suite run --dry-run --json` 只预览 planned/skipped case，不创建 evidence，也不执行 automation。\n`suite run` 会顺序执行 suite 中已有 automation、机器 readiness 已满足且不需要 `manual_check` 的 case，并在最后聚合 `suite report`。\n缺 env、automation env 或 fixture 的 case 默认会跳过；确实要强制执行时，加 `--include-not-ready`。\n确认前置条件后，才用 `--include-manual-check` 执行这类 case。\n所有 case 执行完并写入最终 `result.json` 后，\n`suite report` 会读取各 case evidence 目录并汇总为 `pass`、`fail`、`blocked`、`env_issue`、\n`flaky`、`incomplete` 等状态。`pass` 必须声明已经收集 case 的全部 required evidence；\n否则 suite 会保持 `incomplete`，避免把缺证据的运行误判成通过。\n\n### Test Plan\n\n```bash\nbin/lbs test plan pipeline-debug-chat\nbin/lbs test plan pipeline-debug-chat --json\n```\n\n### Test Start\n\n```bash\nbin/lbs test start pipeline-debug-chat\nbin/lbs test start pipeline-debug-chat --json\n```\n\n`test start` 用于 agent 开始一次浏览器测试前记录 run id、开始时间和推荐 report 命令。\n它会把 `--since \"\"` 写进后续报告命令，减少历史日志污染本次判断。\n如果 case 绑定了自动化脚本，输出里也会包含 `test run` 命令和 evidence 目录。\n\n### Test Automation\n\n```bash\nbin/lbs test run webui-login-state --dry-run\nbin/lbs test run pipeline-debug-chat --dry-run\nbin/lbs test run webui-login-state --run-id login-smoke --output reports/evidence/login-smoke\nbin/lbs test run pipeline-debug-chat --run-id pipeline-smoke --output reports/evidence/pipeline-smoke\n```\n\n查看当前所有带自动化脚本的 case：\n\n```bash\nbin/lbs case list --automation\nbin/lbs case list --json --automation\n```\n\n当前自动化覆盖包括登录态、通用 Pipeline Debug Chat、local-agent runner 的基础回复、\nPromptPreProcessing、RAG、plugin tool、MCP stdio tool、非流式、多模态和 RAG+多模态路径。\n不要在文档里手工维护静态 case 清单；以 `case list --automation` 和 suite 定义为准。\n\n自动化脚本位于 `scripts/e2e/`。它们会保存：\n\n- `console.log`\n- `network.log`\n- `screenshot.png`\n- `automation-result.json`\n\n新增 Debug Chat 类自动化时，优先复用 `scripts/e2e/lib/debug-chat.mjs` 中的 pipeline 打开、\nprompt 发送、visible response leaf 判断和失败信号分类，不要在新脚本里复制 DOM 扫描逻辑。\n\n脚本需要本地安装 Playwright 后才能真正执行：\n\n```bash\nnpm install\nnpx playwright install chromium\n```\n\n`pipeline-debug-chat` 通用自动化建议配置 `LANGBOT_PIPELINE_URL`。如果没有 direct URL，\n脚本会尝试通过 `LANGBOT_PIPELINE_NAME` 从 Pipelines 页面进入目标 pipeline。两者都没有时，\n该自动化会返回 `blocked`，不会伪造通过。\n\nRunner 专用 case 不应复用通用 pipeline 变量。Local Agent、Codex Runner 和\nClaude Code Runner 这类 case 会通过 `automation_pipeline_url_env` /\n`automation_pipeline_name_env` 映射到 case-specific env，例如\n`LANGBOT_LOCAL_AGENT_PIPELINE_URL`。这些 case 如果缺少专用变量，会返回 `blocked`，\n不会退回到 `LANGBOT_PIPELINE_URL`，避免跑错 pipeline 后产生假阳性。\n如果 case 声明了 `setup_automation`，只有 `bin/lbs test run ` 的真实执行路径会先运行这些 setup；\n`test plan`、`suite plan`、`case list` 和 `--dry-run` 只展示它们，不会修改本地环境。\nsetup 可以是 `case:` 或仓库内 `node:scripts/... --flag`，每一步证据会写到主 evidence 目录下的\n`setup/` 子目录。setup 失败时主 automation 不会继续执行；setup 写入 `.env.local` 后，主 automation\n会重新读取环境。用 `setup_provides_env` 声明 setup 会生成的变量，可以让 readiness 正确显示机器可准备状态。\n如果 Debug Chat case 需要固定流式/非流式路径，可以在 case 中设置\n`automation_stream_output: \"1\"` 或 `\"0\"`，脚本会在发送消息前切换 Debug Chat 的 Stream 控件。\n如果 case 需要上传图片，可以设置 `automation_image_base64_fixture` 指向仓库内的 base64 PNG fixture，\n脚本会在 evidence 目录写出临时 PNG 并通过 Debug Chat 上传控件发送。\n`bin/lbs test plan --json` 和 `bin/lbs suite plan --json`\n都会显示这些专用变量是否已配置。\n\n### Test Report 和日志守卫\n\n```bash\nbin/lbs test report pipeline-debug-chat\nbin/lbs test report pipeline-debug-chat --output reports/pipeline-debug-chat.md\nbin/lbs test report pipeline-debug-chat \\\n --backend-log /path/to/backend.log \\\n --frontend-log /path/to/frontend.log \\\n --console-log /path/to/console.log\nbin/lbs test report pipeline-debug-chat --evidence-dir reports/evidence/pipeline-smoke\nbin/lbs test report pipeline-debug-chat --backend-log /path/to/backend.log --json\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs test report pipeline-debug-chat --tail-lines 2000\nbin/lbs test report pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\" --tail-lines 2000\n```\n\n`test report` 会生成报告模板，并默认从 `LANGBOT_REPO/data/logs/` 自动选择最新的\n`langbot-*.log` 作为 LangBot 后端日志扫描。也可以用 `--backend-log` 覆盖，或用\n`--no-auto-log` 只生成模板。\n\n如果提供 `--evidence-dir`，或 `--console-log` 指向 `reports/evidence//console.log`，\n报告会优先读取同目录的 `automation-result.json`，并展示自动化脚本的 `status`、`reason`、\n起止时间和目标 URL。\n\n日志守卫会扫描常见错误、secret 泄露风险、case 声明的 success/failure patterns，以及已知\ntroubleshooting pattern。它不控制浏览器，也不替代 UI 通过判断。`success_patterns`\n命中会作为通过证据写入 `success_signals`；声明了 success pattern 但本次扫描窗口没有命中，\n会给 warning；`failure_patterns` 命中会让本次日志守卫 fail。\n\n建议在执行浏览器 case 前记录当前时间，然后在报告阶段使用 `--since`。如果只想快速看\n最近日志，可以使用 `--tail-lines`。\n\n### Runtime Log Guard\n\n如果还没有进入某个具体 UI case，只是想观察 LangBot 后端日志，可以直接使用 `log`\n命令。它和 `test report` 使用同一套扫描器、secret 脱敏、troubleshooting pattern 和\ncase success/failure pattern。\n\n```bash\nbin/lbs log scan --tail-lines 300\nbin/lbs log scan --case pipeline-debug-chat --since \"2026-05-21T10:30:00+08:00\"\nbin/lbs log scan --backend-log /path/to/langbot.log --json\nbin/lbs log scan --failure-pattern \"runner.tool_loop_error|Action invoke_llm_stream call timed out\" --strict\n```\n\n`log scan` 默认从 `LANGBOT_REPO/data/logs/` 自动选择最新的 `langbot-*.log`。传入\n`--case ` 后，会额外应用该 case 声明的 `success_patterns`、`failure_patterns`\n和 related troubleshooting。默认用于观察，返回码保持 0；加 `--strict` 后，`fail` 或\n`env_issue` 会返回非 0，适合脚本门禁。\n\n运行期观察可以用 `watch`：\n\n```bash\nbin/lbs log watch --case pipeline-debug-chat\nbin/lbs log watch --backend-log /path/to/langbot.log --interval-ms 500\nbin/lbs log watch --duration-ms 30000 --strict --json\n```\n\n`log watch` 默认从启动时的文件末尾开始，只观察新追加的日志；加 `--from-start` 可从文件开头扫。\n它会实时打印新命中的 findings 和 success signals。为了避免当前历史日志噪声影响观察，默认不因\n异常返回非 0；加 `--strict` 后，退出时如果看到 `fail` 或 `env_issue` 会返回非 0。\n\n给一次 QA 运行包日志窗口时，用 `guard start/stop`：\n\n```bash\nbin/lbs log guard start --run-id local-debug --case pipeline-debug-chat\n# 执行浏览器或手工测试\nbin/lbs log guard stop --run-id local-debug\n```\n\n`start` 会在 `reports/log-guards/.json` 记录起始时间、case 和当前后端日志路径；\n`stop` 会用 start/stop 时间作为扫描窗口，生成 `reports/log-guards/.md`，并默认按\nstrict guard 返回码处理。临时只想收集报告、不想让命令失败，可以加 `--no-strict`。\n\n当前 LangBot core 日志还不是完全结构化日志，runtime guard 主要依赖时间窗口和文本 pattern。\n已支持 ISO 时间戳和 LangBot 当前的 `[MM-DD HH:mm:ss.SSS]` 前缀；没有时间戳的连续行会跟随上一条\n带时间戳的日志块。如果后续 core 能提供稳定 request id、conversation id、plugin action id 或\nJSON log field，guard 可以从“时间窗口 + 文本匹配”升级为更精确的关联分析。\n\n### Test Result\n\n```bash\nbin/lbs test result pipeline-debug-chat \\\n --result pass \\\n --reason \"Debug Chat returned OK and the report log guard was clean.\" \\\n --evidence-dir reports/evidence/pipeline-smoke \\\n --started-at \"2026-05-21T10:30:00+08:00\" \\\n --evidence ui,screenshot,console,backend_log\n```\n\n`test result` 用于把一次人工/agent browser 运行的最终判断写成标准 `result.json`，\n供 `suite report` 聚合。它不会替代 UI 测试：如果写 `--result pass`，`--evidence`\n必须覆盖该 case 的 `evidence_required`，否则命令会失败。自动化脚本写\n`automation-result.json`；如果 case 还要求 backend log、API diagnostic 或 filesystem\nevidence，agent 需要在报告和诊断完成后再用 `test result` 写最终 `result.json`。\n\n### Troubleshooting\n\n```bash\nbin/lbs trouble list langbot-testing\nbin/lbs trouble show plugin-runtime-timeout\nbin/lbs trouble search runtime\nbin/lbs trouble add langbot-testing --title \"...\" --symptom \"...\" --cause \"...\" --fix \"...\"\n```\n\n## 目录说明\n\n```text\nskills/\n .env # 共享默认变量\n langbot-env-setup/ # 环境、浏览器、登录态、代理\n langbot-testing/ # WebUI / provider / pipeline 测试\nschemas/ # 结构化资产 schema\nsrc/ # lbs TypeScript 源码\nbin/ # lbs 入口\ndocs/ # 设计文档和用户手册\nAGENTS.md # agent 维护协议\n```\n\n## 添加一个新测试路径\n\n1. 先让 agent 通过浏览器探索并执行路径。\n2. 稳定后创建 case：\n\n```bash\nbin/lbs case new provider-xxx --title \"Provider XXX can be configured\" --area provider --type provider\n```\n\n3. 编辑生成的 `cases/*.yaml`，补充真实步骤、检查点和 troubleshooting。\n\n4. 校验：\n\n```bash\nbin/lbs validate\nbin/lbs index --check\nbin/lbs index\n```\n\n## 添加一个新故障\n\n```bash\nbin/lbs trouble add langbot-testing \\\n --title \"Plugin runtime actions time out\" \\\n --symptom \"Debug Chat shows Agent runner temporarily unavailable\" \\\n --cause \"Old plugin runtime survived backend restart\" \\\n --fix \"Stop old runtime processes and restart LangBot\"\n```\n\n然后编辑生成的 YAML，补充 `patterns`、`related_cases` 和验证方式。\n\n## 当前边界\n\n- `lbs test plan` 只生成测试计划，不直接控制浏览器。\n- `lbs test report` 生成报告，默认扫描最新 LangBot 后端日志；也可扫描显式提供的\n backend/frontend/console 日志文件。\n- 真正的 UI 操作由当前 agent 的浏览器能力执行。\n- `env doctor` 是 readiness check，不是产品测试。\n- `curl/API` 是诊断工具，不是主要测试路径。\n\n\n# LangBot Skills\n\nThis directory is the **single source of truth** for LangBot's agent skills —\nreusable, on-demand instruction packs for AI agents (Claude Code, Codex, Cursor,\nand LangBot's own Local Agent) working with the LangBot ecosystem.\n\n> These skills were consolidated here from the former `langbot-app/langbot-skills`\n> repository (now archived). Documentation and the landing page link here; do not\n> re-copy skill content elsewhere — link to this directory instead.\n\n## Skill catalog\n\n| Skill | What it covers |\n| --- | --- |\n| [`langbot-dev`](skills/langbot-dev) | Core backend + web frontend development (Quart, Vite, API, migrations, MCP server) |\n| [`langbot-plugin-dev`](skills/langbot-plugin-dev) | Plugin SDK / component development, debugging, WebSocket testing |\n| [`langbot-deploy`](skills/langbot-deploy) | Docker / Compose / Kubernetes deployment, config.yaml, Box runtime, global API key |\n| [`langbot-testing`](skills/langbot-testing) | WebUI / e2e QA harness, cases, fixtures, troubleshooting (the `bin/lbs` CLI) |\n| [`langbot-env-setup`](skills/langbot-env-setup) | Local dev/test environment, browser access, OAuth, proxy, startup |\n| [`langbot-mcp-ops`](skills/langbot-mcp-ops) | Operating a LangBot instance through its MCP server (`/mcp`) |\n| [`langbot-space-ops`](skills/langbot-space-ops) | Browsing the LangBot Space marketplaces through the Space MCP server |\n| [`langbot-eba-adapter-dev`](skills/langbot-eba-adapter-dev) | Building platform adapters for the Event-Based Agents architecture |\n| [`langbot-skills-maintenance`](skills/langbot-skills-maintenance) | Adding, deduplicating, and auditing skills in this directory |\n\n`skills.index.json` is the machine-readable index (regenerate with `bin/lbs index`).\n\n## Quick start (for an AI agent)\n\n1. Read this README, `AGENTS.md`, and `docs/user-guide.md` to understand the layout.\n2. Read `skills/.env` for shared local defaults. On a new machine, copy\n `skills/.env.example` to `skills/.env.local` (gitignored) and override\n machine-specific values there. Never commit secrets.\n3. Pick the smallest relevant skill from the catalog above and follow its\n `SKILL.md`.\n\n## The `lbs` CLI\n\nThe testing assets ship with a small CLI (`bin/lbs`, Node >= 22.6). The\n`bin/lbs` wrapper is a generated local entrypoint; on a fresh checkout, run\n`npm run bootstrap` once if it is missing. `npm install` also regenerates it via\nthe `prepare` script.\n\n```bash\nnpm run bootstrap # create bin/lbs if missing\nbin/lbs validate # validate skills/cases/troubleshooting structure\nbin/lbs index # regenerate skills.index.json\nbin/lbs env show # inspect resolved env defaults (redacted)\nbin/lbs env doctor # diagnose local environment readiness\nbin/lbs case list --ready\nbin/lbs test plan \nbin/lbs suite plan langbot-debug-chat-load-gate\n```\n\n## Maintenance rule\n\nWhen the LangBot / LangBot Space **API or MCP server changes**, the\ncorresponding skill here MUST be updated in the same change. The MCP tool\nsurface, the API, and these skills are kept in lockstep — see each repo's\n`AGENTS.md`.\n\n\n{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/case.schema.json\",\n \"title\": \"LangBot Skill Test Case\",\n \"type\": \"object\",\n \"required\": [\n \"id\",\n \"title\",\n \"mode\",\n \"area\",\n \"type\",\n \"priority\",\n \"risk\",\n \"ci_eligible\",\n \"tags\",\n \"skills\",\n \"steps\",\n \"checks\",\n \"evidence_required\"\n ],\n \"allOf\": [\n {\n \"if\": {\n \"properties\": {\n \"mode\": { \"const\": \"agent-browser\" }\n }\n },\n \"then\": {\n \"required\": [\"env\"]\n }\n }\n ],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"mode\": {\n \"type\": \"string\",\n \"enum\": [\"agent-browser\", \"probe\"]\n },\n \"area\": {\n \"type\": \"string\"\n },\n \"type\": {\n \"type\": \"string\",\n \"enum\": [\n \"smoke\",\n \"regression\",\n \"feature\",\n \"provider\",\n \"exploratory\",\n \"contract\",\n \"performance\",\n \"reliability\",\n \"chaos\",\n \"security\"\n ]\n },\n \"priority\": {\n \"type\": \"string\",\n \"enum\": [\"p0\", \"p1\", \"p2\"]\n },\n \"risk\": {\n \"type\": \"string\",\n \"enum\": [\"low\", \"medium\", \"high\"]\n },\n \"ci_eligible\": {\n \"type\": \"boolean\"\n },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"skills\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"env_any\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*(\\\\|[A-Z][A-Z0-9_]*)+$\"\n }\n },\n \"steps\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"checks\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"evidence_required\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"enum\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"backend_log\",\n \"frontend_log\",\n \"api_diagnostic\",\n \"filesystem\",\n \"metrics\",\n \"trace\",\n \"profile\",\n \"resource_log\"\n ]\n },\n \"minItems\": 1\n },\n \"preconditions\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup_automation\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^(?:case:[a-z0-9][a-z0-9_-]*|node:scripts/[A-Za-z0-9_./-]+\\\\.(?:mjs|js|ts)(?:\\\\s+--[A-Za-z0-9][A-Za-z0-9_-]*(?:=[A-Za-z0-9_./:@-]+)?)*)$\"\n }\n },\n \"setup_provides_env\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n }\n },\n \"cleanup\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"diagnostics\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation\": {\n \"type\": \"string\"\n },\n \"automation_env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation_env_any\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*(\\\\|[A-Z][A-Z0-9_]*)+$\"\n }\n },\n \"automation_prompt\": {\n \"type\": \"string\"\n },\n \"automation_prompts_json\": {\n \"type\": \"string\"\n },\n \"automation_expected_text\": {\n \"type\": \"string\"\n },\n \"automation_response_timeout_ms\": {\n \"type\": \"string\"\n },\n \"automation_stream_output\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_image_base64_fixture\": {\n \"type\": \"string\"\n },\n \"automation_runner_config_patch_json\": {\n \"type\": \"string\"\n },\n \"automation_restore_runner_config\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_expected_runner_id\": {\n \"type\": \"string\"\n },\n \"automation_reset_debug_chat\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_session_type\": {\n \"type\": \"string\",\n \"enum\": [\"person\", \"group\"]\n },\n \"automation_debug_chat_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_max_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_requests\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_concurrency\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_timeout_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_first_response_p95_ms\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_max_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_error_rate\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_error_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_ok_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_min_provider_fault_count\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_expected_prefix\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_prompt_template\": {\n \"type\": \"string\"\n },\n \"automation_debug_chat_load_stream\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_reset\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_fail_on_final_mismatch\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_require_success\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_debug_chat_load_provider_model_thresholds_json\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_pipeline_name\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_model_name\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fallback_model_names\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_response_text\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_first_token_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_chunk_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_chunk_count\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_first_n\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_every_n\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_models\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fault_status\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_after_first_chunk\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_fake_provider_fail_after_first_chunk_delay_ms\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_fail_after_first_chunk_mode\": {\n \"type\": \"string\",\n \"enum\": [\"disconnect\", \"error_event\"]\n },\n \"automation_fake_provider_fail_after_first_chunk_models\": {\n \"type\": \"string\"\n },\n \"automation_fake_provider_dynamic_response\": {\n \"type\": \"string\",\n \"enum\": [\"0\", \"1\", \"false\", \"true\"]\n },\n \"automation_filesystem_checks_json\": {\n \"type\": \"string\"\n },\n \"metrics_thresholds_json\": {\n \"type\": \"string\"\n },\n \"load_profile_json\": {\n \"type\": \"string\"\n },\n \"fault_model_json\": {\n \"type\": \"string\"\n },\n \"automation_pipeline_url_env\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n },\n \"automation_pipeline_name_env\": {\n \"type\": \"string\",\n \"pattern\": \"^[A-Z][A-Z0-9_]*$\"\n },\n \"success_patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"failure_patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"expected_failures\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"troubleshooting\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n}\n\n\n# Schemas\n\n这个目录存放 LangBot skills 结构化资产的 JSON Schema。\n\n它们不是测试脚本，也不会执行浏览器动作。它们的作用是定义 agent 和维护者后续新增资产时应该遵守的文件结构。\n\n## 文件说明\n\n- `skills//fixtures/fixtures.json`\n 不是 JSON Schema，但由 `bin/lbs validate` 校验。\n 它登记 deterministic fixture 文件、类型和关联 case，供 `bin/lbs fixture check` 做 readiness 检查。\n\n- `case.schema.json`\n 约束 `skills//cases/*.yaml` 的格式。\n Case 描述 agent-browser 或 probe QA 路径，包括前置条件、步骤、检查点、诊断手段和关联故障。\n\n- `suite.schema.json`\n 约束 `skills//suites/*.yaml` 的格式。\n Suite 只组织 case 集合，用于 smoke、regression 或 release gate 等测试入口。\n\n- `troubleshooting.schema.json`\n 约束 `skills//troubleshooting/*.yaml` 的格式。\n Troubleshooting 条目描述症状、日志/错误模式、可能原因、修复步骤和验证信号。\n\n- `skill-index.schema.json`\n 约束生成文件 `skills.index.json` 的格式。\n 这个索引用于让 agent 快速发现已有 skills、references、cases、suites 和 troubleshooting。\n\n- `reports/evidence//result.json`\n 不是 catalog schema，而是执行期最终裁定产物，由 `bin/lbs test result` 写入。\n `suite report` 读取其中的 `status`、`reason`、起止时间和 `evidence_collected`，\n 并用 `evidence_missing` 防止缺证据的 `pass` 被当作完整通过。\n\n- `reports/evidence//automation-result.json`\n 不是 catalog schema，而是浏览器自动化脚本的原始运行结论，供 `bin/lbs test report`\n 展示和推断日志扫描窗口。\n\n## 为什么需要 schemas\n\nSchemas 是基础设施护栏：\n\n- 防止 case、suite 和 troubleshooting 随着增长变得格式混乱\n- 让 `bin/lbs validate` 能发现缺字段和错误结构\n- 为未来编辑器提示和 CI 校验留接口\n- 帮助 agent 新增资产时知道应该写哪些字段\n\n## 当前校验方式\n\n`bin/lbs validate` 做轻量、schema 对齐的校验，不引入额外依赖。它会检查必填字段、\n枚举值、boolean 字段、重复列表项、automation 脚本存在性，以及 case、suite、skill、\ntroubleshooting 之间的交叉引用。这里的 schema 仍是格式契约；如果未来引入正式 JSON\nSchema validator，应继续保持这些本地交叉引用检查。\n\nCase 里的 `env` / `automation_env` 表示所有列出的变量都需要配置。遇到二选一输入时，\n使用 `env_any` / `automation_env_any`，每一项写成 `LANGBOT_PIPELINE_URL|LANGBOT_PIPELINE_NAME`\n这类 one-of 组合，避免 agent 因为只配置了 URL 或 name 其中之一而误判未就绪。\n`setup` 和 `preconditions` 是人工确认项，会让 readiness 进入 `manual_check`；\n`setup_automation` 是 `test run` 可以自动执行的准备步骤，配合 `setup_provides_env`\n声明它会生成的机器变量。\n\n\n{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/skill-index.schema.json\",\n \"title\": \"LangBot Skills Index\",\n \"type\": \"object\",\n \"required\": [\"generated_by\", \"skills\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"generated_by\": {\n \"type\": \"string\"\n },\n \"skills\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\n \"directory\",\n \"name\",\n \"description\",\n \"references\",\n \"cases\",\n \"case_summaries\",\n \"suites\",\n \"suite_summaries\",\n \"fixtures\",\n \"troubleshooting\",\n \"troubleshooting_summaries\"\n ],\n \"additionalProperties\": false,\n \"properties\": {\n \"directory\": { \"type\": \"string\" },\n \"name\": { \"type\": \"string\" },\n \"description\": { \"type\": \"string\" },\n \"references\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"case_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"mode\", \"area\", \"type\", \"priority\", \"risk\", \"ci_eligible\", \"tags\", \"automation\", \"setup_automation\", \"setup_provides_env\", \"evidence_required\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"mode\": { \"type\": \"string\", \"enum\": [\"agent-browser\", \"probe\"] },\n \"area\": { \"type\": \"string\" },\n \"type\": { \"type\": \"string\" },\n \"priority\": { \"type\": \"string\" },\n \"risk\": { \"type\": \"string\" },\n \"ci_eligible\": { \"type\": \"boolean\" },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"automation\": { \"type\": \"string\" },\n \"setup_automation\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"setup_provides_env\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"evidence_required\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"suites\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"suite_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"description\", \"type\", \"priority\", \"tags\", \"cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"description\": { \"type\": \"string\" },\n \"type\": { \"type\": \"string\" },\n \"priority\": { \"type\": \"string\" },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"fixtures\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"kind\", \"path\", \"related_cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"kind\": { \"type\": \"string\" },\n \"path\": { \"type\": \"string\" },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n },\n \"troubleshooting\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n },\n \"troubleshooting_summaries\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"category\", \"related_cases\"],\n \"additionalProperties\": false,\n \"properties\": {\n \"id\": { \"type\": \"string\" },\n \"title\": { \"type\": \"string\" },\n \"category\": { \"type\": \"string\" },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n }\n }\n }\n }\n }\n }\n}\n\n\n{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/suite.schema.json\",\n \"title\": \"LangBot Skill Test Suite\",\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"description\", \"type\", \"priority\", \"tags\", \"cases\"],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"description\": {\n \"type\": \"string\"\n },\n \"type\": {\n \"type\": \"string\",\n \"enum\": [\n \"smoke\",\n \"regression\",\n \"release_gate\",\n \"exploratory\",\n \"contract\",\n \"performance\",\n \"reliability\",\n \"chaos\",\n \"security\"\n ]\n },\n \"priority\": {\n \"type\": \"string\",\n \"enum\": [\"p0\", \"p1\", \"p2\"]\n },\n \"tags\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n }\n }\n}\n\n\n{\n \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n \"$id\": \"https://langbot.app/schemas/langbot-skills/troubleshooting.schema.json\",\n \"title\": \"LangBot Skill Troubleshooting Entry\",\n \"type\": \"object\",\n \"required\": [\"id\", \"title\", \"symptoms\", \"patterns\", \"likely_causes\", \"fix_steps\", \"verification\"],\n \"additionalProperties\": true,\n \"properties\": {\n \"id\": {\n \"type\": \"string\",\n \"pattern\": \"^[a-z0-9][a-z0-9_-]*$\"\n },\n \"title\": {\n \"type\": \"string\"\n },\n \"date\": {\n \"type\": \"string\"\n },\n \"category\": {\n \"type\": \"string\",\n \"enum\": [\"product\", \"env_issue\", \"external_dependency\", \"blocked\", \"flaky\"]\n },\n \"symptoms\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"patterns\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"likely_causes\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"fix_steps\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" },\n \"minItems\": 1\n },\n \"verification\": {\n \"type\": \"string\"\n },\n \"related_cases\": {\n \"type\": \"array\",\n \"items\": { \"type\": \"string\" }\n }\n }\n}\n\n\n{\n \"generated_by\": \"lbs\",\n \"skills\": [\n {\n \"directory\": \"langbot-deploy\",\n \"name\": \"langbot-deploy\",\n \"description\": \"Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \\\"deploy langbot\\\", \\\"langbot docker\\\", \\\"langbot compose\\\", \\\"langbot kubernetes\\\", \\\"langbot config.yaml\\\", \\\"langbot box runtime\\\", \\\"langbot global api key\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-dev\",\n \"name\": \"langbot-dev\",\n \"description\": \"Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \\\"langbot backend\\\", \\\"langbot dev\\\", \\\"langbot api\\\", \\\"add langbot endpoint\\\", \\\"langbot migration\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-eba-adapter-dev\",\n \"name\": \"langbot-eba-adapter-dev\",\n \"description\": \"Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing.\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-env-setup\",\n \"name\": \"langbot-env-setup\",\n \"description\": \"Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths.\",\n \"references\": [\n \"references/browser-access-selection.md\",\n \"references/computer-use.md\",\n \"references/oauth-browser-profile.md\",\n \"references/playwright-mcp.md\",\n \"references/proxy.md\",\n \"references/service-startup.md\",\n \"references/wsl-notes.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-mcp-ops\",\n \"name\": \"langbot-mcp-ops\",\n \"description\": \"Operate a LangBot instance through its built-in MCP (Model Context Protocol) server. Use when an AI agent needs to manage LangBot — list/create/update/delete bots, agents, pipelines, models, knowledge bases, MCP servers, and skills — over MCP instead of raw HTTP. Covers the /mcp endpoint, API-key auth (web-UI lbk_ keys and the config.yaml global key), the tool surface, and client configuration. Triggers on \\\"langbot mcp\\\", \\\"manage langbot via mcp\\\", \\\"langbot /mcp\\\", \\\"langbot mcp server\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-plugin-dev\",\n \"name\": \"langbot-plugin-dev\",\n \"description\": \"Develop, debug, and test LangBot plugins. Use when creating new LangBot plugins, fixing plugin bugs, setting up a LangBot test environment, or testing plugins via WebSocket. Covers plugin component architecture (EventListener, Command, Tool), the plugin SDK API (invoke_llm, get_llm_models, send_message, plugin storage), common pitfalls, and automated WebSocket-based testing. Triggers on \\\"langbot plugin\\\", \\\"lbp\\\", \\\"GroupChatSummary\\\", \\\"plugin debug\\\", \\\"langbot test\\\".\",\n \"references\": [\n \"references/test-env-setup.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-skills-maintenance\",\n \"name\": \"langbot-skills-maintenance\",\n \"description\": \"Maintain the langbot-skills repository with low duplication. Use when adding, editing, or auditing LangBot skills, references, cases, troubleshooting entries, indexes, or periodic entropy-control checks for this skills repository.\",\n \"references\": [\n \"references/curation-workflow.md\"\n ],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-space-ops\",\n \"name\": \"langbot-space-ops\",\n \"description\": \"Browse and search the LangBot Space marketplaces (plugins, MCP servers, skills) through the Space MCP server. Use when an AI agent needs to discover LangBot extensions on space.langbot.app over MCP. Covers the /mcp endpoint, Personal Access Token (PAT) auth, the tool surface, and client configuration. Triggers on \\\"langbot space mcp\\\", \\\"search langbot plugins\\\", \\\"langbot marketplace mcp\\\", \\\"space.langbot.app mcp\\\".\",\n \"references\": [],\n \"cases\": [],\n \"case_summaries\": [],\n \"suites\": [],\n \"suite_summaries\": [],\n \"fixtures\": [],\n \"troubleshooting\": [],\n \"troubleshooting_summaries\": []\n },\n {\n \"directory\": \"langbot-testing\",\n \"name\": \"langbot-testing\",\n \"description\": \"Test LangBot WebUI and core product flows with an automated browser and backend logs. Use when validating the configured LangBot frontend, pipeline Debug Chat, model provider setup and test buttons, bot and knowledge-base UI flows, or troubleshooting failed LangBot end-to-end tests.\",\n \"references\": [\n \"references/agent-runner-qa-workflow.md\",\n \"references/agent-runner-release-gate.md\",\n \"references/dify-agent-runner.md\",\n \"references/langrag-knowledge-base.md\",\n \"references/local-agent-runner-coverage.md\",\n \"references/local-agent-runner.md\",\n \"references/mcp-stdio-testing.md\",\n \"references/model-provider-testing.md\",\n \"references/performance-reliability-testing.md\",\n \"references/pipeline-debug-chat.md\",\n \"references/plugin-e2e-smoke.md\",\n \"references/sandbox-skill-authoring.md\",\n \"references/skill-all-tool-acceptance.md\",\n \"references/troubleshooting.md\",\n \"references/web-ui-testing.md\",\n \"references/workspace-release-testing.md\"\n ],\n \"cases\": [\n \"acp-agent-runner-debug-chat\",\n \"agent-run-ledger-audit\",\n \"agent-runner-async-db-readiness\",\n \"agent-runner-behavior-matrix\",\n \"agent-runner-fixture-contract\",\n \"agent-runner-health-visibility\",\n \"agent-runner-ledger-concurrency\",\n \"agent-runner-ledger-contention\",\n \"agent-runner-ledger-invariants\",\n \"agent-runner-ledger-stress\",\n \"agent-runner-live-install\",\n \"agent-runner-qa-debug-chat\",\n \"agent-runner-release-preflight\",\n \"agent-runner-runtime-chaos\",\n \"bot-event-routing-product-flow\",\n \"box-mcp-heartbeat-recovery\",\n \"claude-code-agent-debug-chat\",\n \"codex-agent-debug-chat\",\n \"dify-agent-debug-chat\",\n \"langbot-fake-provider-debug-chat-cross-pipeline-isolation\",\n \"langbot-fake-provider-debug-chat-fault-recovery\",\n \"langbot-fake-provider-debug-chat-load\",\n \"langbot-fake-provider-debug-chat-slow-load\",\n \"langbot-fault-taxonomy-contract\",\n \"langbot-live-backend-latency\",\n \"langbot-live-backend-log-health\",\n \"langbot-live-control-plane-api\",\n \"langbot-overhead-accounting-contract\",\n \"langbot-space-debug-chat-concurrency-smoke\",\n \"langrag-kb-retrieve\",\n \"langrag-parser-golden-e2e\",\n \"langrag-sentinel-kb-discover\",\n \"local-agent-basic-debug-chat\",\n \"local-agent-combo-rag-compaction-tool-debug-chat\",\n \"local-agent-complex-coding-task-debug-chat\",\n \"local-agent-context-compaction-debug-chat\",\n \"local-agent-effective-prompt-debug-chat\",\n \"local-agent-model-fallback-before-first-chunk-debug-chat\",\n \"local-agent-multimodal-debug-chat\",\n \"local-agent-multitool-rag-compaction-debug-chat\",\n \"local-agent-nonstreaming-debug-chat\",\n \"local-agent-parallel-tools-rag-compaction-debug-chat\",\n \"local-agent-plugin-tool-call-debug-chat\",\n \"local-agent-rag-debug-chat\",\n \"local-agent-rag-multimodal-debug-chat\",\n \"local-agent-steering-debug-chat\",\n \"local-agent-streaming-post-commit-failure-debug-chat\",\n \"local-agent-tool-error-recovery-debug-chat\",\n \"local-agent-tool-loop-limit-debug-chat\",\n \"mcp-stdio-register\",\n \"mcp-stdio-tool-call\",\n \"pipeline-debug-chat\",\n \"pipeline-debug-chat-performance\",\n \"plugin-e2e-smoke\",\n \"provider-deepseek\",\n \"qa-plugin-smoke-live-install\",\n \"sandbox-skill-authoring-e2e\",\n \"sandbox-skill-authoring-edit-existing-e2e\",\n \"skill-discovery-via-mcp-gateway\",\n \"webui-login-state\",\n \"wizard-onebot-agent-runtime\",\n \"wizard-runner-marketplace-catalog\",\n \"wizard-scenario-routing\",\n \"workspace-compatibility-preflight\",\n \"workspace-plugin-pages-smoke\",\n \"workspace-repository-contracts\"\n ],\n \"case_summaries\": [\n {\n \"id\": \"acp-agent-runner-debug-chat\",\n \"title\": \"ACP Runner can answer through Debug Chat using real remote Claude\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"acp\",\n \"claude\",\n \"external-runner\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-acp-agent-runner-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_ACP_RUNNER_PIPELINE_URL\",\n \"LANGBOT_ACP_RUNNER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"agent-run-ledger-audit\",\n \"title\": \"Persisted Runner run ledger passes end-to-end invariants\",\n \"mode\": \"probe\",\n \"area\": \"agent\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"ledger\",\n \"audit\",\n \"tools\"\n ],\n \"automation\": \"scripts/e2e/agent-run-ledger-audit.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\",\n \"metrics\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-async-db-readiness\",\n \"title\": \"Runner async DB readiness probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"async-db\",\n \"aiosqlite\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-async-db-readiness.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-behavior-matrix\",\n \"title\": \"Runner deterministic behavior matrix probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"deterministic-runner\",\n \"protocol\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-behavior-matrix.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-fixture-contract\",\n \"title\": \"QA Runner fixture contract probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"fixture\",\n \"deterministic-runner\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-fixture-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-health-visibility\",\n \"title\": \"Agent configuration shows whether its selected runner is usable\",\n \"mode\": \"agent-browser\",\n \"area\": \"agent\",\n \"type\": \"feature\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"agent\",\n \"productization\",\n \"release-gate\"\n ],\n \"automation\": \"scripts/e2e/agent-runner-health-visibility.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-concurrency\",\n \"title\": \"Runner run ledger concurrency and auth pytest probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"pytest\",\n \"run-ledger\",\n \"concurrency\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-concurrency.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-contention\",\n \"title\": \"Runner ledger SQLite contention probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"stress\",\n \"ledger\",\n \"concurrency\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-contention.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-invariants\",\n \"title\": \"Runner ledger schema and status invariants probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"ledger\",\n \"invariant\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-invariants.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-ledger-stress\",\n \"title\": \"Runner ledger lightweight stress probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"stress\",\n \"ledger\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-ledger-stress.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-live-install\",\n \"title\": \"QA Runner package installs and registers in LangBot\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"plugin\",\n \"local-install\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/install-qa-plugin-smoke.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"agent-runner-qa-debug-chat\",\n \"title\": \"QA Runner returns deterministic output through Debug Chat\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"pipeline\",\n \"debug-chat\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"case:agent-runner-live-install\",\n \"node:scripts/e2e/ensure-qa-agent-runner-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_QA_RUNNER_PIPELINE_URL\",\n \"LANGBOT_QA_RUNNER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-release-preflight\",\n \"title\": \"Agent runner release gate preflight validates environment readiness\",\n \"mode\": \"agent-browser\",\n \"area\": \"release\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"release-gate\",\n \"preflight\",\n \"environment\"\n ],\n \"automation\": \"scripts/e2e/agent-runner-release-preflight.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"agent-runner-runtime-chaos\",\n \"title\": \"Runner SDK runtime chaos pytest probe\",\n \"mode\": \"probe\",\n \"area\": \"release\",\n \"type\": \"regression\",\n \"priority\": \"p0\",\n \"risk\": \"high\",\n \"ci_eligible\": true,\n \"tags\": [\n \"agent-runner\",\n \"probe\",\n \"pytest\",\n \"runtime\",\n \"sdk\"\n ],\n \"automation\": \"skills/langbot-testing/probes/agent-runner-runtime-chaos.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"filesystem\"\n ]\n },\n {\n \"id\": \"bot-event-routing-product-flow\",\n \"title\": \"Bot event routing and adapter input can be inspected from the WebUI\",\n \"mode\": \"agent-browser\",\n \"area\": \"bot\",\n \"type\": \"feature\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"bot\",\n \"eba\",\n \"event-routing\",\n \"productization\"\n ],\n \"automation\": \"scripts/e2e/bot-event-routing-product-flow.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"box-mcp-heartbeat-recovery\",\n \"title\": \"Box heartbeat restores an existing stdio MCP tool session\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"box\",\n \"mcp\",\n \"stdio\",\n \"reliability\",\n \"chaos\",\n \"fault-injection\"\n ],\n \"automation\": \"skills/langbot-testing/probes/box-mcp-heartbeat-recovery.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"metrics\",\n \"api_diagnostic\",\n \"resource_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"claude-code-agent-debug-chat\",\n \"title\": \"Claude Code Agent can retrieve LangBot history through MCP\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"claude-code\",\n \"external-runner\",\n \"mcp\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"codex-agent-debug-chat\",\n \"title\": \"Codex Agent can retrieve LangBot history through MCP\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"agent-runner\",\n \"codex\",\n \"external-runner\",\n \"mcp\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"dify-agent-debug-chat\",\n \"title\": \"Dify Runner returns a response through Pipeline Debug Chat\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"provider\",\n \"priority\": \"p2\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"dify\",\n \"agent-runner\",\n \"pipeline\"\n ],\n \"automation\": \"\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-cross-pipeline-isolation\",\n \"title\": \"LangBot Debug Chat fake-provider cross-pipeline isolation probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"reliability\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"isolation\",\n \"concurrency\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-cross-pipeline-isolation.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-cross-pipelines.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_A_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_A_NAME\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_B_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_B_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-fault-recovery\",\n \"title\": \"LangBot Debug Chat fake-provider fault recovery probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"chaos\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"fault-injection\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-load\",\n \"title\": \"LangBot Debug Chat controlled fake-provider load probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"load\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fake-provider-debug-chat-slow-load\",\n \"title\": \"LangBot Debug Chat slow fake-provider load probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"fake-provider\",\n \"slow-provider\",\n \"load\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_BASE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PID\",\n \"LANGBOT_FAKE_PROVIDER_PROVIDER_UUID\",\n \"LANGBOT_FAKE_PROVIDER_MODEL_UUID\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-fault-taxonomy-contract\",\n \"title\": \"LangBot fault taxonomy and cleanup contract\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": true,\n \"tags\": [\n \"reliability\",\n \"chaos\",\n \"contract\",\n \"synthetic\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-fault-taxonomy-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-backend-latency\",\n \"title\": \"LangBot live backend basic latency probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"live-backend\",\n \"latency\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-backend-latency.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-backend-log-health\",\n \"title\": \"LangBot live backend log health probe\",\n \"mode\": \"probe\",\n \"area\": \"reliability\",\n \"type\": \"reliability\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"reliability\",\n \"live-backend\",\n \"backend-log\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-backend-log-health.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"backend_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-live-control-plane-api\",\n \"title\": \"LangBot live control-plane API probe\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"reliability\",\n \"live-backend\",\n \"control-plane\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-live-control-plane-api.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-overhead-accounting-contract\",\n \"title\": \"LangBot overhead accounting metrics contract\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": true,\n \"tags\": [\n \"performance\",\n \"metrics\",\n \"contract\",\n \"synthetic\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-overhead-accounting-contract.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"metrics\",\n \"resource_log\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langbot-space-debug-chat-concurrency-smoke\",\n \"title\": \"LangBot Debug Chat real Space-provider concurrency smoke\",\n \"mode\": \"probe\",\n \"area\": \"performance\",\n \"type\": \"performance\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"performance\",\n \"debug-chat\",\n \"websocket\",\n \"space\",\n \"live-provider\",\n \"smoke\",\n \"metrics\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_PIPELINE_URL\",\n \"LANGBOT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_MODEL_UUID\",\n \"LANGBOT_E2E_MODEL_UUID\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"langrag-kb-retrieve\",\n \"title\": \"LangRAG knowledge base ingests and retrieves a sentinel document\",\n \"mode\": \"agent-browser\",\n \"area\": \"knowledge\",\n \"type\": \"feature\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"langrag\",\n \"knowledge\",\n \"rag\"\n ],\n \"automation\": \"scripts/e2e/langrag-kb-retrieve.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"langrag-parser-golden-e2e\",\n \"title\": \"LangRAG and GeneralParsers retrieve a structured golden document\",\n \"mode\": \"agent-browser\",\n \"area\": \"knowledge\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"golden\",\n \"e2e\",\n \"langrag\",\n \"parser\",\n \"knowledge\"\n ],\n \"automation\": \"\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"langrag-sentinel-kb-discover\",\n \"title\": \"Existing LangRAG sentinel knowledge base is discoverable\",\n \"mode\": \"probe\",\n \"area\": \"knowledge\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"langrag\",\n \"knowledge\",\n \"rag\",\n \"fixture\"\n ],\n \"automation\": \"scripts/e2e/ensure-langrag-sentinel-kb.mjs\",\n \"setup_automation\": [],\n \"setup_provides_env\": [],\n \"evidence_required\": [\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-basic-debug-chat\",\n \"title\": \"Local Agent Debug Chat returns a deterministic streaming response\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"smoke\",\n \"priority\": \"p0\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"streaming\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-combo-rag-compaction-tool-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and plugin tool result in one Debug Chat run\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-complex-coding-task-debug-chat\",\n \"title\": \"Local Agent completes a multi-file coding task with iterative verification\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"agentic\",\n \"coding\",\n \"sandbox\",\n \"tools\",\n \"long-running\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/reset-complex-agent-task.mjs\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_COMPLEX_AGENT_WORKSPACE\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\",\n \"metrics\"\n ]\n },\n {\n \"id\": \"local-agent-context-compaction-debug-chat\",\n \"title\": \"Local Agent compacts long Debug Chat history and preserves older facts\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"context\",\n \"compaction\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-effective-prompt-debug-chat\",\n \"title\": \"Local Agent consumes host effective prompt after PromptPreProcessing\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"prompt\",\n \"plugin\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-model-fallback-before-first-chunk-debug-chat\",\n \"title\": \"Local Agent falls back when the primary model fails before streaming starts\",\n \"mode\": \"probe\",\n \"area\": \"pipeline\",\n \"type\": \"chaos\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"fallback\",\n \"streaming\",\n \"fake-provider\",\n \"fault-injection\"\n ],\n \"automation\": \"skills/langbot-testing/probes/langbot-debug-chat-concurrency.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-fake-provider-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_FAKE_PROVIDER_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_URL\",\n \"LANGBOT_FAKE_PROVIDER_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"metrics\",\n \"network\",\n \"api_diagnostic\",\n \"filesystem\"\n ]\n },\n {\n \"id\": \"local-agent-multimodal-debug-chat\",\n \"title\": \"Local Agent Debug Chat preserves uploaded image input\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"multimodal\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"network\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-multitool-rag-compaction-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and two-step plugin tool loop\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"tool-loop\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-nonstreaming-debug-chat\",\n \"title\": \"Local Agent Debug Chat returns a deterministic response with UI streaming disabled\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"medium\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"pipeline\",\n \"non-streaming\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\"\n ]\n },\n {\n \"id\": \"local-agent-parallel-tools-rag-compaction-debug-chat\",\n \"title\": \"Local Agent preserves RAG, compacted history, and parallel plugin tools\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"context\",\n \"compaction\",\n \"plugin\",\n \"tools\",\n \"parallel-tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-plugin-tool-call-debug-chat\",\n \"title\": \"Local Agent can call a plugin-provided tool\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"plugin\",\n \"tools\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"case:qa-plugin-smoke-live-install\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-rag-debug-chat\",\n \"title\": \"Local Agent Debug Chat answers from a LangRAG knowledge base\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p1\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"langrag\",\n \"pipeline\"\n ],\n \"automation\": \"scripts/e2e/pipeline-debug-chat.mjs\",\n \"setup_automation\": [\n \"node:scripts/e2e/ensure-local-agent-pipeline.mjs --write-env\",\n \"node:scripts/e2e/ensure-langrag-sentinel-kb.mjs --write-env\"\n ],\n \"setup_provides_env\": [\n \"LANGBOT_LOCAL_AGENT_PIPELINE_URL\",\n \"LANGBOT_LOCAL_AGENT_PIPELINE_NAME\",\n \"LANGBOT_LOCAL_AGENT_RAG_KB_UUID\"\n ],\n \"evidence_required\": [\n \"ui\",\n \"screenshot\",\n \"console\",\n \"backend_log\",\n \"api_diagnostic\"\n ]\n },\n {\n \"id\": \"local-agent-rag-multimodal-debug-chat\",\n \"title\": \"Local Agent preserves image input while using RAG context\",\n \"mode\": \"agent-browser\",\n \"area\": \"pipeline\",\n \"type\": \"regression\",\n \"priority\": \"p2\",\n \"risk\": \"high\",\n \"ci_eligible\": false,\n \"tags\": [\n \"local-agent\",\n \"rag\",\n \n\n\n\n\n---\nname: langbot-deploy\ndescription: Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on \"deploy langbot\", \"langbot docker\", \"langbot compose\", \"langbot kubernetes\", \"langbot config.yaml\", \"langbot box runtime\", \"langbot global api key\".\n---\n\n# LangBot Deployment & Configuration\n\nCovers running LangBot in production. For development see `langbot-dev`.\n\nRead https://langbot.app/llms.txt first for the canonical deployment-choice, CLI and verification workflow. The summary below applies when selecting hosting.\n\n## Deployment choice and optional managed hosting\n\nRespect the user's deployment preference and reuse an existing instance when available. Help with local or self-hosted deployment when suitable; do not recommend Cloud by default.\n\nOffer [LangBot Cloud](https://space.langbot.app/cloud) when the user wants managed hosting without maintaining a server, the local environment is unsuitable, or a deployment problem remains difficult after reasonable troubleshooting. Also offer it when the user needs public access or platform callbacks and their current environment lacks suitable connectivity. Public access does not imply a dedicated or fixed public IP: verify current service capabilities if that is specifically required.\n\nIf the user wants to support LangBot's continued development and open-source maintenance, mention the official hosted service as an optional way to do so, without pressure or repeated appeals. Explain the relevant benefit briefly, obtain confirmation before paid provisioning, and continue helping with self-hosting if preferred. Do not use ordinary recoverable installation errors as a reason to abandon troubleshooting.\n\n## Docker Compose (recommended)\n\n```bash\ngit clone https://github.com/langbot-app/LangBot\ncd LangBot/docker\n\n# Full stack (sandbox/Box + stdio MCP hosting + skill add/edit enabled)\ndocker compose --profile all up\n\n# Basic (no Box runtime)\ndocker compose up\n```\n\nThe `all` / `box` profile starts three services:\n\n- `langbot` — main app, serves API + UI on `:5300`.\n- `langbot_plugin_runtime` — plugin runtime (control `:5400`, debug `:5401`).\n- `langbot_box` — Box sandbox runtime (`:5410`). Uses the host Docker socket to\n spawn sandbox containers, so the **Box root host path and in-container path\n must be identical** (`BOX__LOCAL__HOST_ROOT=${LANGBOT_BOX_ROOT:-${PWD}/data/box}`).\n OSS allows its RPC and managed-process relay to run without a token when both\n sides leave `LANGBOT_BOX_CONTROL_TOKEN` unset. For an exposed endpoint, set\n the same value of at least 32 non-whitespace characters in both the LangBot\n and Box containers. Generate it once with `openssl rand -hex 32`; never put\n it in `box.runtime.endpoint` or commit it to config.\n\nA Compose deployment may optionally set\n`LANGBOT_PLUGIN_RUNTIME_CONTROL_TOKEN` on both `langbot` and\n`langbot_plugin_runtime` when port 5400 needs shared-secret protection. OSS\ndefaults to leaving it unset on both sides. If enabled, generate one value with\n`openssl rand -hex 32`; configuring only one side causes the control connection\nto fail. Kubernetes may use the `langbot-plugin-runtime-control` Secret shown in\n`docker/kubernetes.yaml`.\n\nWith Box off, the dashboard/skills list stays visible (read-only) but sandbox\ntools, skill add/edit, and stdio MCP are disabled. Set `box.enabled: false`\n(or `BOX__ENABLED=false`) to match.\n\n## Kubernetes\n\nSee `docker/kubernetes.yaml` and the deployment guide at\nhttps://langbot.app/docs. `docker/deploy-k8s-test.sh` is a test helper.\n\n## config.yaml (generated at `data/config.yaml` on first run)\n\nTop-level sections: `api`, `system`, `command`, `concurrency`, `proxy`,\n`database`, `vdb`, `storage`, `plugin`, `monitoring`, `box`, `space`.\n\nKey settings:\n\n| Key | Meaning |\n| --- | --- |\n| `api.port` | HTTP API + UI port (default 5300) |\n| `api.global_api_key` | **Global API key** for the HTTP API + MCP server. Non-empty = accepted with no login/DB record; no `lbk_` prefix required. Empty = disabled. Plaintext — trusted/internal only, serve over HTTPS. |\n| `plugin.runtime_ws_url` | Standalone plugin runtime WS URL (e.g. `ws://langbot_plugin_runtime:5400/control/ws`) |\n| `box.enabled` | Master switch for the Box sandbox runtime |\n| `box.backend` | `local` (Docker/nsjail autopick) / `docker` / `nsjail` / `e2b` / explicit unsafe `host`; env override `BOX__BACKEND` |\n| `box.runtime.endpoint` | External Box runtime URL (e.g. `ws://127.0.0.1:5410`); empty = local auto-managed |\n\nMany keys have `ENV__SUBKEY` overrides (e.g. `BOX__BACKEND`, `BOX__ENABLED`).\n\n## Runtimes & flags\n\n- LangBot started directly spawns the plugin runtime over **stdio**.\n- In containers it connects to a standalone runtime over **WebSocket**; start\n with `--standalone-runtime`.\n- Box has a parallel `--standalone-box` flag; the Docker box host is\n `langbot_box:5410`.\n- `box.backend: host` runs commands directly as the Box Runtime system user.\n It is never auto-selected, provides no sandbox isolation, and is only for\n trusted local development. A WebSocket-controlled host backend requires\n `LANGBOT_BOX_CONTROL_TOKEN`; local stdio control is allowed.\n\n## Global API key — enabling for agents/automation\n\n```yaml\n# data/config.yaml\napi:\n port: 5300\n global_api_key: 'a-strong-secret' # empty disables it\n```\n\nThis key authenticates both the HTTP API and the MCP server (`/mcp`) without a\nlogin session. See `langbot-mcp-ops` for using it, and `docs/API_KEY_AUTH.md`.\n\n## Pitfalls\n\n- \"No supported sandbox backend (Docker / nsjail / E2B)\" with Docker running\n usually means the user isn't in the `docker` group →\n `sudo usermod -aG docker ` and restart in a new shell.\n- Do not use `box.backend: host` as a production fallback. It cannot enforce\n image, filesystem, network, PID, CPU, memory, or storage isolation.\n- Box root host/container path mismatch breaks sandbox container creation.\n- Don't commit a non-empty `api.global_api_key` to version control.\n\n\n---\nname: langbot-dev\ndescription: Develop, build, and debug the LangBot core backend and web frontend. Use when working inside the LangBot repository — backend (Python/Quart, src/langbot/pkg), the Vite/React web UI, HTTP API controllers/services, Alembic migrations, or the MCP server. Covers the dev environment (uv, pnpm), repo layout, the API auth model (user token / API key / global key), adding API endpoints, and the rule that API changes must update the MCP server and skills. Triggers on \"langbot backend\", \"langbot dev\", \"langbot api\", \"add langbot endpoint\", \"langbot migration\".\n---\n\n# LangBot Core Development\n\nThis skill covers developing the LangBot core (the main repo), distinct from\nplugin development (see `langbot-plugin-dev`) and deployment (`langbot-deploy`).\n\n## Stack\n\n- **Backend**: Python `>=3.11,<4.0`, deps via `uv`. Framework: **Quart** (async\n Flask). Serves the HTTP API + pre-built web UI on `http://127.0.0.1:5300`.\n- **Frontend** (`web/`): **Vite + React Router 7 + shadcn/ui + Tailwind**,\n managed by `pnpm`. Dev server on `:3000`. (NOT Next.js — `dev` script is `vite`.)\n\n## Dev environment\n\n```bash\n# Backend\npip install uv\nuv sync --dev\nuv run main.py # API + UI on http://127.0.0.1:5300\n\n# Frontend (separate terminal)\ncd web\ncp .env.example .env\npnpm install\npnpm dev # http://127.0.0.1:3000 (reads VITE_API_BASE_URL)\n\n# Lint/format hooks (CI runs the same checks)\nuv run pre-commit install\n```\n\nFirst run generates `data/config.yaml`; DB defaults to SQLite (PostgreSQL\nsupported). Migrations run automatically on startup.\n\n## Repo layout (key paths)\n\n```\nsrc/langbot/\n├── __main__.py # entrypoint, CLI flags (--standalone-runtime/-box/--debug)\n├── pkg/\n│ ├── api/\n│ │ ├── http/ # Quart controllers + services\n│ │ │ ├── controller/groups/ # route groups (@group.group_class)\n│ │ │ └── service/ # business logic (called by controllers AND MCP)\n│ │ └── mcp/ # MCP server (server.py = tools, mount.py = ASGI dispatch)\n│ ├── core/ # app bootstrap, stages, task manager\n│ ├── platform/ provider/ pipeline/ plugin/ box/ skill/ rag/ vector/\n│ ├── command/ persistence/ storage/ config/ entity/ telemetry/\n│ └── templates/config.yaml # config template (top-level: api, system, plugin, box, space...)\n├── web/ # Vite SPA\n└── docker/ # compose deployment\n```\n\n## HTTP API auth model\n\nRoute auth is declared per-route via `AuthType` in\n`pkg/api/http/controller/group.py`:\n\n- `NONE` — public.\n- `USER_TOKEN` — web UI JWT (`Authorization: Bearer `).\n- `API_KEY` — `X-API-Key` or `Authorization: Bearer `.\n- `USER_TOKEN_OR_API_KEY` — either.\n\nAuthenticated routes receive an immutable `RequestContext` containing the\nprincipal, authorized Workspace membership, fixed-role permissions, instance,\nrequest id, and placement generation. A browser's `X-Workspace-Id` is only a\nselector and is always checked against the Account membership. Tenant services\nmust accept this context (or an explicit trusted execution context) and fail\nclosed when it is absent.\n\nAPI-key authentication accepts:\n\n1. the **global key** from `config.yaml` `api.global_api_key` only for a\n community instance with exactly one local Workspace, then\n2. **web-UI keys** whose one-time `lbk_` secret is stored only as a hash and is\n bound to one Workspace, explicit scopes, status, and optional expiry.\n\nAn API key derives its Workspace from the key record and ignores a caller's\nWorkspace selector. Public Bot/Webhook routes similarly derive Workspace from\nthe opaque owning resource rather than a header.\n\nRoute groups self-register via `@group.group_class(name, path)` and are\ndiscovered by `importutil.import_modules_in_pkg`.\n\n## Adding an API endpoint\n\n1. Add/extend a controller in `pkg/api/http/controller/groups/` and the matching\n service method in `pkg/api/http/service/`.\n2. Pick the right `AuthType`.\n3. **If the endpoint should be agent-accessible, add/adjust the matching MCP tool\n in `pkg/api/mcp/server.py` and update the `langbot-mcp-ops` skill.** API and\n MCP surface must stay aligned (see `AGENTS.md`).\n4. If `lbctl` uses the route or capability, check the separate\n [`langbot-cli`](https://github.com/langbot-app/langbot-cli) client for compatibility.\n5. Update `docs/service-api-openapi.json` if you maintain the OpenAPI overview.\n\n## Database migrations (Alembic)\n\nSingle migration set supports SQLite + PostgreSQL. Files in\n`src/langbot/pkg/persistence/alembic/versions/`.\n\n```bash\n# From project root (needs data/config.yaml)\nuv run python -m langbot.pkg.persistence.alembic_runner autogenerate \"description\"\n```\n\n## Standards\n\n- All code comments/docstrings in **English**; user-facing strings need **i18n**\n (`en_US` + `zh_Hans` minimum, `ja_JP` where present).\n- Consider toC and toB compatibility + security.\n- Commit format: `(): ` (feat/fix/docs/refactor/...).\n\n## Tests\n\n```bash\nuv run pytest tests/unit_tests -q # unit tests\nuv run pytest tests/unit_tests/api -q # API service tests\nuv run python tests/manual/mcp_smoke.py # MCP server e2e smoke\n```\n\n## See also\n\n- `langbot-plugin-dev` — plugin SDK / runtime development.\n- `langbot-testing` — WebUI/e2e QA harness (`bin/lbs`).\n- `langbot-deploy` — Docker/compose deployment + config.\n- `langbot-mcp-ops` — operating the LangBot MCP server.\n- [`langbot-cli`](https://github.com/langbot-app/langbot-cli) — `lbctl`, a standalone Service API client for managing running Workspaces.\n\n\n---\nname: langbot-eba-adapter-dev\ndescription: Build, refactor, and test LangBot platform adapters for the Event-Based Agents architecture. Use when adding or migrating Telegram, Discord, or other messaging platform adapters to the EBA adapter layout, validating unified event/message conversion, writing live adapter probes, or using standalone plugin runtime plus Computer Use for end-to-end platform testing.\n---\n\n# LangBot EBA Adapter Development\n\nUse this skill when implementing or reviewing a LangBot platform adapter under the Event-Based Agents architecture.\n\n## Controlling a running instance via MCP\n\nBeyond writing code, you can **drive a live LangBot instance over MCP** — no raw\nHTTP needed. Two MCP servers exist (both reuse existing API keys; see `AGENTS.md`):\n\n- **LangBot instance** — `http://:5300/mcp` (auth: web-UI `lbk_` key or the\n `api.global_api_key` from `config.yaml`). Manage bots, pipelines, models,\n knowledge bases, and skills. See the **`langbot-mcp-ops`** skill.\n- **LangBot Space marketplace** — `https://space.langbot.app/mcp` (auth: Personal\n Access Token). Search plugins / MCP servers / skills. See the\n **`langbot-space-ops`** skill.\n\n> Any change to an agent-accessible HTTP API endpoint must keep the matching MCP\n> tool and these skills in sync.\n\n## Core Rule\n\nDo not let platform-native event or message shapes leak into LangBot's common path. Each adapter must convert incoming SDK objects into unified EBA entities before dispatch:\n\n- Events: `langbot_plugin.api.entities.builtin.platform.events`\n- Message chains: `langbot_plugin.api.entities.builtin.platform.message.MessageChain`\n- Users/groups/members: `langbot_plugin.api.entities.builtin.platform.entities`\n- Raw platform objects may remain only in `source_platform_object` for debugging or platform-specific escape hatches.\n\n## Start Here\n\n1. Read the EBA design docs in `LangBot/docs/event-based-agents/`.\n2. Read the architecture-level acceptance checklist before writing or validating code:\n - `LangBot/docs/event-based-agents/adapters/acceptance-checklist.md`\n3. Read the current reference adapter before writing code. Prefer Telegram first:\n - `LangBot/src/langbot/pkg/platform/adapters/telegram/`\n - `LangBot/docs/event-based-agents/adapters/telegram.md`\n4. Read the legacy source adapter for the target platform:\n - `LangBot/src/langbot/pkg/platform/sources/.py`\n - `LangBot/src/langbot/pkg/platform/sources/.yaml`\n5. Inspect SDK entity definitions in `langbot-plugin-sdk/src/langbot_plugin/api/entities/builtin/platform/`.\n6. Search before assuming APIs. Platform SDKs change often.\n\n## Adapter Layout\n\nCreate one directory per adapter:\n\n```text\nLangBot/src/langbot/pkg/platform/adapters//\n├── __init__.py\n├── adapter.py\n├── api_impl.py\n├── event_converter.py\n├── manifest.yaml\n├── message_converter.py\n├── platform_api.py\n├── types.py\n└── .svg\n```\n\nAdd optional helpers such as `voice.py` only when the platform has a real domain-specific surface.\n\nEnsure `pyproject.toml` package data includes adapter assets:\n\n```toml\npackage-data = { \"langbot\" = [\"templates/**\", \"pkg/platform/sources/*\", \"pkg/platform/adapters/**\", ...] }\n```\n\n## Implementation Checklist\n\n- `manifest.yaml` declares `metadata.name`, config schema, supported events, common APIs, and platform-specific APIs.\n- `adapter.py` creates the platform client, subscribes to native events, filters self/bot loops where appropriate, calls `event_converter.target2yiri(...)`, then dispatches the EBA event.\n- `event_converter.py` maps native events to EBA event classes such as `MessageReceivedEvent`, `MessageEditedEvent`, `MessageDeletedEvent`, `MessageReactionEvent`, `MemberJoinedEvent`, `BotInvitedToGroupEvent`, and `PlatformSpecificEvent`.\n- `message_converter.py` maps native messages to `MessageChain`, and maps `MessageChain` back to the platform send format.\n- `api_impl.py` implements common EBA APIs: send, reply, edit, delete, forward, user/group/member lookup, moderation, upload/file URL, leave group.\n- `platform_api.py` keeps platform-specific calls behind `call_platform_api(action, params)`.\n- Unsupported common APIs must raise explicit SDK platform errors such as `NotSupportedError`; do not silently no-op.\n- Destructive APIs such as kick, ban, leave, delete, or moderation must be gated in live tests and documented.\n\n## Conversion Contract\n\nFor message events, the common shape should look like this regardless of platform:\n\n```python\nplatform_events.MessageReceivedEvent(\n type=\"message.received\",\n adapter_name=\"\",\n message_id=,\n message_chain=platform_message.MessageChain([...]),\n sender=platform_entities.User(...),\n chat_type=platform_entities.ChatType.PRIVATE or ChatType.GROUP,\n chat_id=,\n group=platform_entities.UserGroup(...) or None,\n source_platform_object=,\n)\n```\n\nMessage content should use common components:\n\n- `Source` for original message id/time when available.\n- `Plain` for text.\n- `At` / `AtAll` for mentions.\n- `Image`, `Voice`, `File` for media.\n- `Forward` only when the platform can represent or emulate it safely.\n\nIf a platform event cannot cleanly map to a common event, emit `PlatformSpecificEvent` with a compact `action` and structured `data`.\n\n## Unit Tests\n\nAdd focused tests under `LangBot/tests/unit_tests/platform/test__eba_adapter.py`.\n\nCover at least:\n\n- Manifest supported events match adapter `supported_events()`.\n- Manifest supported APIs match adapter `supported_apis()`.\n- Platform API map matches manifest actions.\n- Dispatcher chooses the most specific EBA listener.\n- Message converter maps every supported common component both directions where possible:\n - `Source`\n - `Plain`\n - `At`\n - `AtAll`\n - `Image`\n - `Voice`\n - `File`\n - `Quote`\n - `Face`\n - `Forward`\n - `Unknown`\n - mixed chains preserving order\n- Event converter maps message received/edited/deleted/reaction, raw uncached gateway events, member events, and bot join/leave events.\n- Send/reply methods pass correct platform kwargs and return `MessageResult`.\n\nRun the existing reference adapter tests too:\n\n```bash\ncd LangBot\nuv run pytest tests/unit_tests/platform/test__eba_adapter.py tests/unit_tests/platform/test_telegram_eba_adapter.py\nuv run python -m py_compile tests/e2e/live__eba_probe.py\ngit diff --check\n```\n\n## Live Test Workflow\n\nDirect adapter live probes are useful diagnostics, but they are not sufficient acceptance evidence for EBA. Treat `tests/e2e/live__eba_probe.py` as an auxiliary tool only. The final adapter record must distinguish:\n\n- `plugin-e2e-ui`: real SDK plugin through standalone runtime, LangBot core, adapter, and a real/simulator UI action. This can mark an inbound UI item complete.\n- `plugin-e2e-protocol`: real SDK plugin through standalone runtime, LangBot core, adapter, and a protocol-boundary injected event. This is useful evidence but must not be claimed as UI coverage.\n- `plugin-e2e-outbound`: real SDK plugin calls an API and the bot output is visible in the real/simulator UI. This can mark send/API coverage complete.\n- `adapter-live`: direct adapter probe connected to a real/simulator endpoint. This is auxiliary only.\n- `unit`: mocked conversion/API-shape coverage. This is auxiliary only.\n- `not-supported`: platform protocol or SDK has no equivalent. Must include the reason.\n- `blocked`: intended capability could not be verified. This is not complete.\n\nWrite a live probe in `LangBot/tests/e2e/live__eba_probe.py`. It should:\n\n1. Read token/client ids from environment variables or CLI args.\n2. Start the adapter directly.\n3. Register an EBA listener and write JSONL evidence to `LangBot/data/temp/`.\n4. Wait for a real user/platform event instead of fabricating the entrypoint.\n5. Exercise common APIs and `call_platform_api` actions.\n6. Observe returned gateway events for edit/delete/reaction/member/bot lifecycle where available.\n7. Print a summary containing passed, failed, skipped, and observed event types.\n8. Redact or avoid printing secrets.\n9. Keep destructive operations behind flags and run them last.\n\nUse Computer Use when the user asks for real platform end-to-end coverage. Actually send messages/click reactions in the platform UI or otherwise trigger real user-side events; do not replace that with unit tests.\n\nFor media/component acceptance, keep the direction and trigger source explicit:\n\n- Real inbound media only counts when a human-side platform UI or simulator UI sends the image/file/voice to the bot and the plugin JSONL records the corresponding common component.\n- Bot outbound media only proves `send_message`/adapter send conversion. It does not prove inbound conversion.\n- Protocol-boundary injection, such as sending a OneBot event directly into a reverse WebSocket adapter, is useful and should be labelled `plugin-e2e-protocol`, but it must not be reported as UI-level end-to-end media upload.\n- If the UI cannot send or upload the media, record the item as `blocked` with the exact client/simulator limitation.\n\n## Standalone Runtime + Plugin Test\n\nWhen validating the whole LangBot EBA path, test with the SDK standalone runtime and a real test plugin. This is the required acceptance path; direct adapter calls do not prove the EBA architecture path.\n\nThe required path is:\n\n```text\nReal platform / simulator UI\n -> platform SDK native event\n -> adapter event converter\n -> unified EBA event/entity/message types\n -> LangBot core event dispatch\n -> standalone SDK runtime\n -> real test plugin listener\n -> plugin calls platform APIs through SDK\n -> LangBot core API dispatch\n -> adapter API implementation\n -> real platform / simulator UI\n```\n\nTypical shape:\n\n```bash\n# Terminal 1, SDK repo\ncd langbot-plugin-sdk\nuv run python -m langbot_plugin.cli.__init__ rt \\\n --debug-only \\\n --ws-control-port 5400 \\\n --ws-debug-port 5401 \\\n --skip-deps-check\n\n# Terminal 2, LangBot repo\ncd LangBot\nexport PYTHONPATH=/absolute/path/to/langbot-plugin-sdk/src:${PYTHONPATH:-}\nuv run main.py --standalone-runtime\n\n# Terminal 3, plugin directory\nexport DEBUG_RUNTIME_WS_URL=ws://127.0.0.1:5401/plugin/ws\nexport EBA_PROBE_LOG=/absolute/path/to/LangBot/data/temp/_eba_plugin_probe.jsonl\nexport EBA_PROBE_API=1\nexport EBA_PROBE_COMPONENT_SWEEP=1\nexport EBA_PROBE_PLATFORM_API=1\nuv --project /absolute/path/to/langbot-plugin-sdk run python -m langbot_plugin.cli.__init__ run\n```\n\nUse an EBA probe plugin that subscribes to all relevant EBA event classes and runs SDK API calls after the first `MessageReceived`.\n\nThe plugin evidence should be JSONL and include:\n\n- event class and `event.type`\n- adapter name\n- chat type and chat ID\n- sender/user/group IDs with secrets redacted\n- `bot_uuid` and `adapter_name`, proving LangBot filled common routing fields before plugin dispatch\n- received `message_chain` component list\n- API action name, input summary, result or error\n- unsupported or blocked reason when an item is skipped\n\nFor full adapter acceptance, enable both probe sweeps:\n\n- `EBA_PROBE_COMPONENT_SWEEP=1` sends the required outbound message components through `send_message`.\n- `EBA_PROBE_PLATFORM_API=1` calls common safe APIs plus selected `call_platform_api` actions for the adapter.\n\nThe SDK must support `plugin.call_platform_api(bot_uuid, action, params)` for platform-specific acceptance. If the SDK cannot call a platform-specific action from the plugin, the adapter cannot be fully accepted even if direct adapter probes pass.\n\n## Required EBA Acceptance Coverage\n\nBefore marking an adapter migrated, fill out an adapter record against `LangBot/docs/event-based-agents/adapters/acceptance-checklist.md`.\n\nAt minimum, the record must cover these categories:\n\n- Message receive component tests through `plugin-e2e-ui`: `Source`, `Plain`, `At`, `AtAll`, `Image`, `Voice`, `File`, `Quote`, `Face`, `Forward`, `Unknown`, and mixed chains where the platform supports them. Protocol-only receive evidence must be labelled `plugin-e2e-protocol`.\n- Message send component tests through `plugin-e2e-outbound`: `Plain`, `At`, `AtAll`, `Image`, `Voice`, `File`, `Quote`, `Face`, `Forward`, and mixed chains where the platform supports them.\n- Every event declared in `manifest.yaml -> spec.supported_events`.\n- Every common API declared in `manifest.yaml -> spec.supported_apis.required` and `optional`.\n- Every action declared in `manifest.yaml -> spec.platform_specific_apis`.\n- Compatibility tests for manifest declarations, legacy message listener fallback, EBA listener specificity, bot self-message filtering, and `source_platform_object` reply/debug behavior.\n\nDo not declare an event or API in the manifest unless it has an implementation path and an acceptance entry. If a platform or simulator lacks a capability, document it as `not-supported` or `blocked` rather than silently omitting the test.\n\n## Common Pitfalls\n\n- `get_bots()` may return bot dictionaries, not UUID strings. Probe plugins should select an enabled dict and pass `bot[\"uuid\"]` to `get_bot_info()` and `send_message()`.\n- Make sure the probe subscribes to every event you claim to verify. Missing `MessageDeleted` subscription can make a working adapter look untested.\n- Some platforms emit both cached and raw gateway events, producing duplicate evidence for delete/reaction. Count this explicitly; do not treat duplicates as failure unless semantics differ.\n- Self-message filtering is platform-specific. Filter bot-originated `message.received` loops, but do not accidentally filter edit/delete events needed for bot-owned API probes.\n- Reaction events may be filtered for bot self reactions. To test user reaction add/remove, use real UI interaction or a real user token path if permitted.\n- File uploads usually happen as message attachments. A standalone `upload_file` API may need to be `NotSupportedError`.\n- Live probes should not leak bot tokens through command output, logs, docs, or final answers.\n- Discord requires privileged intents for message content and members. Missing intents can look like converter bugs.\n- Telegram Bot API exposes only limited member lists; document capability gaps.\n- Do not mark moderation APIs verified unless they ran against a disposable target member/bot.\n- If `leave_group` is tested, run it last because the test bot will be removed from the server/group.\n- Restore local LangBot DB/test state after live runs if you enabled temporary bots or changed plugin settings.\n\n## Documentation Record\n\nAdd or update `LangBot/docs/event-based-agents/adapters/.md` in the same style as Telegram:\n\n- Status and adapter directory.\n- Configuration table matching manifest fields.\n- Supported EBA event list.\n- Common API table with support and limitations.\n- `call_platform_api` action list.\n- Receive component table with evidence level per component.\n- Send component table with evidence level per component.\n- Event table with evidence level per event.\n- Common API table with evidence level per API.\n- Platform-specific API table with evidence level per action.\n- Live test record with exact date, endpoint/simulator, standalone runtime command, test plugin path/name, JSONL evidence path, channel/group type, observed events, APIs exercised, destructive operations, and skipped items.\n\nBe honest. Put untested or skipped APIs in the document with the reason. Do not imply full parity when a platform cannot provide the same information density.\n\n## Before Finishing\n\n- Run unit tests and compile the live probe.\n- Run the standalone runtime plugin E2E path for every required acceptance item that the platform supports.\n- Run `git diff --check`.\n- Summarize live JSONL evidence by event type.\n- Stop all long-running runtimes and probes.\n- Confirm no secrets are staged.\n- Leave unrelated untracked files alone.\n\n\n# Browser Access Selection\n\nChoose the lightest browser-control path that can complete the task.\n\n## Decision Order\n\n1. If Codex Computer Use, Claude Computer Use, or another visible browser-control tool is available, use `computer-use.md`.\n2. If no computer-control tool is available but Playwright MCP is available, use `playwright-mcp.md`.\n3. If the browser session must survive restarts or OAuth login is required, also use `oauth-browser-profile.md`.\n4. If running under WSL, add `wsl-notes.md`.\n5. If external sites or model providers time out, add `proxy.md`.\n\n## Principle\n\nComputer Use and Playwright MCP are alternative browser-control paths. Both still need LangBot services to be reachable, so service checks stay in `service-startup.md`.\n\n\n# Computer Use Browser Path\n\nUse this path when Codex Computer Use, Claude Computer Use, or another agent-visible browser-control capability is available.\n\n## Why This Path Is Simpler\n\nComputer Use can interact with a visible browser directly, so it usually does not need Playwright MCP configuration or a separate MCP browser bridge.\n\n## Workflow\n\n1. Verify LangBot backend/frontend with `service-startup.md`.\n2. Open the WebUI in the controlled browser.\n3. If login is needed, let the user complete GitHub OAuth. Never handle credentials or 2FA.\n4. Keep the browser/profile available for later testing.\n5. Hand off to `langbot-testing` after the page shows the logged-in WebUI.\n\n## Still Required\n\n- Proxy may still be needed for GitHub OAuth or model provider tests. Use `proxy.md`.\n- Persisted profile details may still matter if the computer-control browser is restarted. Use `oauth-browser-profile.md` if login state must survive.\n\n\n# OAuth Browser Profile\n\nUse this reference when LangBot or LangBot Space needs GitHub OAuth login and the agent must reuse the authenticated browser state later.\n\nRead `skills/.env` first for `LANGBOT_BACKEND_URL`, `LANGBOT_FRONTEND_URL`, `LANGBOT_BROWSER_PROFILE`, `LANGBOT_CHROMIUM_EXECUTABLE`, and proxy defaults.\n\n## Rules\n\n- Never handle the user's GitHub password, passkey, recovery code, or 2FA secret.\n- Open a visible browser and let the user complete credential steps.\n- Reuse a fixed browser profile path.\n- Do not print token values. It is acceptable to report localStorage key names.\n\n## Manual Visible Login Flow\n\n1. Verify LangBot backend is reachable with `service-startup.md`.\n2. Launch a visible Chromium window with the persistent profile:\n\n```bash\nsetsid \"$LANGBOT_CHROMIUM_EXECUTABLE\" \\\n --no-sandbox \\\n --ozone-platform=x11 \\\n --user-data-dir=\"$LANGBOT_BROWSER_PROFILE\" \\\n --proxy-server=\"$LANGBOT_PROXY_SOCKS\" \\\n --proxy-bypass-list=\"$LANGBOT_NO_PROXY\" \\\n \"$LANGBOT_BACKEND_URL/login\" \\\n >/tmp/langbot-visible-chrome.log 2>&1 < /dev/null &\n```\n\n3. The user completes:\n\n```text\nLogin with Space -> Login with GitHub -> GitHub credentials / 2FA -> authorize\n```\n\n4. The agent can then reuse the same profile for automated checks.\n\n## Expected Successful State\n\nAfter login, LangBot should redirect away from `/login`, for example to a `/home/...` URL on the selected origin.\n\nExpected visible signals:\n\n```text\nLangBot\nDashboard\nHome\nBots\nPipelines\nKnowledge\nExtensions\n```\n\nExpected localStorage key names:\n\n```text\ntoken\nuserEmail\nlangbot_language\n```\n\nIf the user logged in on one origin but `LANGBOT_FRONTEND_URL` still shows `/login`, copy only the auth state needed between origins. Do not print token values.\n\n\n# Playwright MCP Browser Path\n\nUse this path when the agent needs browser automation but no Computer Use browser-control path is available.\n\n## Known Paths\n\n- Persistent browser profile: `LANGBOT_BROWSER_PROFILE` from `skills/.env.local`\n- Chromium executable: `LANGBOT_CHROMIUM_EXECUTABLE` from `skills/.env.local`\n- Codex MCP config: `$CODEX_HOME/config.toml` or the config path used by the active agent.\n\n## MCP Config\n\nKeep the profile path fixed so the agent can reuse authenticated state.\n\n```toml\n[mcp_servers.playwright]\ncommand = \"npx\"\nargs = [\"-y\", \"@playwright/mcp@latest\", \"--no-sandbox\", \"--executable-path\", \"\", \"--proxy-server\", \"\", \"--proxy-bypass\", \"localhost,127.0.0.1\", \"--user-data-dir\", \"\"]\n```\n\nAfter changing MCP config, restart Codex so the MCP server is relaunched with the new args.\n\n## Visible Login\n\nFor OAuth login, Playwright MCP's headless browser is not enough. Launch a visible browser with the same profile and let the user complete login. Use `oauth-browser-profile.md`.\n\n## Common Failures\n\n- MCP still uses old args after editing config: restart Codex or kill old `playwright-mcp` processes and restart the session.\n- Browser is headless during OAuth: use the visible login command from `oauth-browser-profile.md`.\n\n\n# Proxy Setup\n\nUse this reference when GitHub OAuth, package installation, model provider tests, or external API calls time out.\n\nRead defaults from `skills/.env` first.\n\n## Standard Local Proxy\n\n```bash\nexport HTTP_PROXY=\"$LANGBOT_PROXY_HTTP\"\nexport HTTPS_PROXY=\"$LANGBOT_PROXY_HTTP\"\nexport ALL_PROXY=\"$LANGBOT_PROXY_SOCKS\"\nexport http_proxy=\"$LANGBOT_PROXY_HTTP\"\nexport https_proxy=\"$LANGBOT_PROXY_HTTP\"\nexport all_proxy=\"$LANGBOT_PROXY_SOCKS\"\nexport NO_PROXY=\"$LANGBOT_NO_PROXY\"\nexport no_proxy=\"$LANGBOT_NO_PROXY\"\n```\n\n## Rule\n\nKeep uppercase and lowercase proxy variables consistent. Different libraries read different names.\n\n## Checks\n\n```bash\nenv | rg -i '^(http|https|all|no)_?proxy='\ncurl -I --max-time 8 --proxy \"$LANGBOT_PROXY_SOCKS\" https://github.com\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL\"\n```\n\n\n# Service Startup\n\nUse this reference for LangBot backend/frontend readiness checks regardless of OS or browser-control method. Read `skills/.env` first and override those defaults with user-provided values or detected running services.\n\n## Variables\n\n- `LANGBOT_REPO`\n- `LANGBOT_WEB_REPO`\n- `LANGBOT_BACKEND_URL`\n- `LANGBOT_FRONTEND_URL`\n- `LANGBOT_DEV_FRONTEND_URL`\n\n## Backend\n\nStart LangBot from the backend repo:\n\n```bash\ncd \"$LANGBOT_REPO\"\nuv run main.py\n```\n\nHealthy startup includes:\n\n```text\nRunning on http://0.0.0.0:\nConnected to plugin runtime.\nPlugin langbot-team/LocalAgent initialized\n```\n\nQuick check:\n\n```bash\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL/login\"\n```\n\nIf `bin/lbs env doctor` reports that `LANGBOT_BACKEND_URL` has no TCP listener,\nthe backend is not running at the configured host and port. A reachable\nstandalone frontend on `LANGBOT_FRONTEND_URL` does not prove backend readiness.\n\nPrefer a visible terminal session while debugging backend startup. Detached\nbackground startup methods can hide early process exits in local agent runs; if\nyou use one, immediately verify both the process and the listener:\n\n```bash\nps -eo pid,cmd | rg 'main.py|uv run main|langbot'\nss -ltnp | rg ':5300'\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL/login\"\n```\n\n## Frontend\n\nStart the new frontend from the web repo:\n\n```bash\ncd \"$LANGBOT_WEB_REPO\"\nVITE_API_BASE_URL=\"$LANGBOT_BACKEND_URL\" pnpm dev --host 0.0.0.0\n```\n\nHealthy startup includes:\n\n```text\nLocal: \n```\n\nQuick check:\n\n```bash\ncurl -I --max-time 3 \"$LANGBOT_FRONTEND_URL\"\n```\n\nIf `VITE_API_BASE_URL` is missing, Vite still serves the page but frontend API\ncalls may go to the frontend port instead of the backend port. That produces\nfalse browser failures in login, wizard, pipeline, and Debug Chat cases.\n\n## Completion Signal\n\nEnvironment setup is not complete until the required frontend/backend URLs are reachable and the chosen browser-control path can open the WebUI.\n\n\n# WSL Notes\n\nUse this reference only for WSL-specific details. Do not put generic LangBot startup or browser-login steps here.\n\n## Network\n\nGitHub login and model provider calls may require proxy access from WSL.\n\nWorking proxy form:\n\n```bash\nsocks5://127.0.0.1:7890\n```\n\nBypass local LangBot:\n\n```bash\nlocalhost,127.0.0.1\n```\n\nQuick checks:\n\n```bash\ncurl -I --max-time 8 --proxy socks5h://127.0.0.1:7890 https://github.com\ncurl -I --max-time 3 \"$LANGBOT_BACKEND_URL\"\n```\n\n## Visible Browser\n\nIf OAuth requires a visible browser, WSL must have a usable display path. If a visible Chromium launch fails, check the local WSL GUI/X11 setup before changing LangBot config.\n\n## Common Failures\n\n- `ERR_NETWORK_CHANGED` or GitHub timeout: browser is not using the SOCKS proxy.\n- LangBot connection refused: backend is not running or not reachable from WSL.\n- User cannot type credentials: browser is headless or not visible to the user.\n\n\n---\nname: langbot-env-setup\ndescription: Prepare a local LangBot development and testing environment for an AI agent. Use when setting up WSL or Linux development, shared local URL variables, proxy variables, backend/frontend startup, Playwright MCP browser access, GitHub OAuth browser login, persisted Chrome profiles, or future Codex computer-use environment paths.\n---\n\n# LangBot Environment Setup\n\nUse this skill when a task needs LangBot to be in a testable state before product testing or development verification.\n\n## Routing\n\n- **Shared local variables**: read `../.env` before using URL, path, browser profile, or proxy defaults.\n- **Always start here**: read `references/browser-access-selection.md` to choose the browser-control path.\n- **LangBot service checks and startup**: read `references/service-startup.md`.\n- **Computer Use available**: read `references/computer-use.md`. This path usually needs less browser/MCP setup.\n- **No Computer Use, browser automation required**: read `references/playwright-mcp.md`.\n- **GitHub OAuth or persisted login profile**: read `references/oauth-browser-profile.md`.\n- **WSL-specific notes**: read `references/wsl-notes.md` only when running under WSL.\n- **Proxy setup**: read `references/proxy.md` when external login, model provider tests, or package downloads time out.\n- **Headless-only automation**: use only after a profile already contains a valid LangBot login. Do not ask the agent to enter GitHub credentials or 2FA.\n\n## Rules\n\n- Never handle the user's GitHub password, passkey, recovery code, or 2FA secret.\n- For OAuth login, open a visible browser and let the user complete the credential steps.\n- Reuse a fixed browser profile path so the agent can later access the logged-in LangBot session.\n- Keep environment-specific paths and commands in `references/`, not in this file.\n- Treat environment setup as complete only after the target LangBot services are reachable and the browser profile can access the WebUI.\n\n\n"}