{"owner":"thewh1teagle","repo":"vibe","hasSkills":true,"totalSkillsCount":1,"totalTokensCount":367,"categories":["root-instruction"],"hasMcp":false,"mcpConfig":null,"found":["AGENTS.md"],"skills":{"AGENTS.md":"# Claude Development Notes\n\n## Tooling\n\n- Tasks: `chore` — see `chore list`\n- JS: `pnpm` only; Python: `uv` standalone scripts (`uv run`)\n\n## Validation\n\nPlan validation scripts: `plans/<name>/<name>_001.py` (+ `.md`), standalone `uv` scripts.\n\n## Skills\n\nCustom skills live in `.claude/skills/<name>/SKILL.md`.\n\n## Execution Mindset\n\nAgent mode: parallel moves, instant iteration, speed by default. Split heavy work until obvious. Estimate by output size — 300 lines = minutes. Correct over easy: rework is slower than doing it right once. Plan resolved → execute, don't re-analyze.\n\n## Working in parallel\n\nSplit large work across ~4–5 subagents, one disjoint module-group each — no shared files; only the coordinator touches shared/root config. Fix contracts (interfaces, design docs) up front. Throughput is per-agent (~100 tok/s, no global limit), so N agents ≈ N× speed — delegate bulk implementation to strong subagents (e.g. Opus 5 for correctness-sensitive code); keep the main loop for planning, contracts, and integration.\n\n## ETA rule\n\nQuote minutes, never days: single tasks 1–10 min, multi-agent work tens of minutes.\n\n```\nminutes ≈ (LOC × 40) / (6000 × N_agents) + ~2 min per stage\n```\n\n## File size\n\n700 lines max. On hitting it, ask before splitting.\n\nSplit by responsibility into halves — find where the file does two jobs and move one out whole. Not a line-count cut, not a `utils` skim. Keep the public API where callers expect it; move tests with their code.\n"},"files":{"AGENTS.md":"# Claude Development Notes\n\n## Tooling\n\n- Tasks: `chore` — see `chore list`\n- JS: `pnpm` only; Python: `uv` standalone scripts (`uv run`)\n\n## Validation\n\nPlan validation scripts: `plans/<name>/<name>_001.py` (+ `.md`), standalone `uv` scripts.\n\n## Skills\n\nCustom skills live in `.claude/skills/<name>/SKILL.md`.\n\n## Execution Mindset\n\nAgent mode: parallel moves, instant iteration, speed by default. Split heavy work until obvious. Estimate by output size — 300 lines = minutes. Correct over easy: rework is slower than doing it right once. Plan resolved → execute, don't re-analyze.\n\n## Working in parallel\n\nSplit large work across ~4–5 subagents, one disjoint module-group each — no shared files; only the coordinator touches shared/root config. Fix contracts (interfaces, design docs) up front. Throughput is per-agent (~100 tok/s, no global limit), so N agents ≈ N× speed — delegate bulk implementation to strong subagents (e.g. Opus 5 for correctness-sensitive code); keep the main loop for planning, contracts, and integration.\n\n## ETA rule\n\nQuote minutes, never days: single tasks 1–10 min, multi-agent work tens of minutes.\n\n```\nminutes ≈ (LOC × 40) / (6000 × N_agents) + ~2 min per stage\n```\n\n## File size\n\n700 lines max. On hitting it, ask before splitting.\n\nSplit by responsibility into halves — find where the file does two jobs and move one out whole. Not a line-count cut, not a `utils` skim. Keep the public API where callers expect it; move tests with their code.\n"},"items":[{"name":"AGENTS.md","path":"AGENTS.md","rawUrl":"https://raw.githubusercontent.com/thewh1teagle/vibe/HEAD/AGENTS.md","title":"AI Agent Protocol & Instructions","category":"root-instruction","format":"markdown","content":"# Claude Development Notes\n\n## Tooling\n\n- Tasks: `chore` — see `chore list`\n- JS: `pnpm` only; Python: `uv` standalone scripts (`uv run`)\n\n## Validation\n\nPlan validation scripts: `plans/<name>/<name>_001.py` (+ `.md`), standalone `uv` scripts.\n\n## Skills\n\nCustom skills live in `.claude/skills/<name>/SKILL.md`.\n\n## Execution Mindset\n\nAgent mode: parallel moves, instant iteration, speed by default. Split heavy work until obvious. Estimate by output size — 300 lines = minutes. Correct over easy: rework is slower than doing it right once. Plan resolved → execute, don't re-analyze.\n\n## Working in parallel\n\nSplit large work across ~4–5 subagents, one disjoint module-group each — no shared files; only the coordinator touches shared/root config. Fix contracts (interfaces, design docs) up front. Throughput is per-agent (~100 tok/s, no global limit), so N agents ≈ N× speed — delegate bulk implementation to strong subagents (e.g. Opus 5 for correctness-sensitive code); keep the main loop for planning, contracts, and integration.\n\n## ETA rule\n\nQuote minutes, never days: single tasks 1–10 min, multi-agent work tens of minutes.\n\n```\nminutes ≈ (LOC × 40) / (6000 × N_agents) + ~2 min per stage\n```\n\n## File size\n\n700 lines max. On hitting it, ask before splitting.\n\nSplit by responsibility into halves — find where the file does two jobs and move one out whole. Not a line-count cut, not a `utils` skim. Keep the public API where callers expect it; move tests with their code.\n","isInternal":false,"tokens":367,"sizeBytes":1503}],"systemPromptSnippet":"<yakaai_skills repo=\"thewh1teagle/vibe\">\n<!-- File: AGENTS.md (Tokens: ~367 | Category: root-instruction) -->\n# Claude Development Notes\n\n## Tooling\n\n- Tasks: `chore` — see `chore list`\n- JS: `pnpm` only; Python: `uv` standalone scripts (`uv run`)\n\n## Validation\n\nPlan validation scripts: `plans/<name>/<name>_001.py` (+ `.md`), standalone `uv` scripts.\n\n## Skills\n\nCustom skills live in `.claude/skills/<name>/SKILL.md`.\n\n## Execution Mindset\n\nAgent mode: parallel moves, instant iteration, speed by default. Split heavy work until obvious. Estimate by output size — 300 lines = minutes. Correct over easy: rework is slower than doing it right once. Plan resolved → execute, don't re-analyze.\n\n## Working in parallel\n\nSplit large work across ~4–5 subagents, one disjoint module-group each — no shared files; only the coordinator touches shared/root config. Fix contracts (interfaces, design docs) up front. Throughput is per-agent (~100 tok/s, no global limit), so N agents ≈ N× speed — delegate bulk implementation to strong subagents (e.g. Opus 5 for correctness-sensitive code); keep the main loop for planning, contracts, and integration.\n\n## ETA rule\n\nQuote minutes, never days: single tasks 1–10 min, multi-agent work tens of minutes.\n\n```\nminutes ≈ (LOC × 40) / (6000 × N_agents) + ~2 min per stage\n```\n\n## File size\n\n700 lines max. On hitting it, ask before splitting.\n\nSplit by responsibility into halves — find where the file does two jobs and move one out whole. Not a line-count cut, not a `utils` skim. Keep the public API where callers expect it; move tests with their code.\n</yakaai_skills>"}