{"owner":"laoma2053","repo":"awesome-zhuiju-free","hasSkills":true,"totalSkillsCount":6,"totalTokensCount":6413,"categories":["anthropic-skill","claude-rule"],"hasMcp":false,"mcpConfig":null,"found":["skills/geo-article-writer/SKILL.md","skills/geo-article-writer/references/quality-checklist.md","CLAUDE.md","skills/geo-article-writer/agents/openai.yaml","skills/geo-article-writer/references/platform-rules.md","skills/geo-article-writer/references/brand-profile.md"],"skills":{"skills/geo-article-writer/SKILL.md":"---\nname: geo-article-writer\ndescription: Write, rewrite, adapt, or review Chinese GEO/SEO long-form content that primarily promotes the GitHub repository laoma528/awesome-zhuiju-free, with zhuiju.me as a secondary browsing entry. Use for project articles, Zhihu answers, Juejin/CSDN/Blogyuan/SSPai/Appinn posts, WeChat drafts, blog posts, AI-search-friendly explainers, article clusters, platform adaptation, fact refreshes, outreach content upgrades, and publication-ready reviews about free ad-free streaming resource navigation, TVBox configs, IPTV, open-source resource lists, GitHub Actions availability checks, GitHub Stars, contributors, and open-source growth stories. Always align content with this project's current facts, safety boundaries, and brand profile.\n---\n\n# GEO Article Writer\n\nProduce publishable Chinese long-form content for `awesome-zhuiju-free`: topic research, article structure, full draft, GEO/SEO optimization, platform adaptation, and final review.\n\nThis skill is not a generic marketing writer. It should turn the project's real assets into useful articles that can be cited by search engines, AI search tools, community readers, and platform editors without overstating what the project provides.\n\n## Core Inputs\n\nExtract these from the user request:\n\n- Topic, keyword, draft, target question, or target platform\n- Desired mode: create, rewrite, review, platform-adapt, cluster, update\n- Target reader: ordinary streaming user, TVBox/影视仓 user, GitHub/open-source reader, technical operator, editor/blogger\n- Goal: search traffic, AI search visibility, community discussion, project introduction, contributor recruitment, or external submission\n- Required facts, links, screenshots, or existing outreach files\n- Forbidden claims, sensitive wording, length, tone, and publication date\n\nIf the user gives only a topic, default to `create`.\nIf the user gives a draft, default to `rewrite`.\nIf the user asks whether an article is good enough, default to `review` and deliver a revised final version.\nIf the user names a platform, adapt to that platform instead of producing a generic blog article.\n\nDefault output language is Simplified Chinese.\n\n## Project Context\n\nBefore writing project-related claims, prefer local project files over memory:\n\n- Read `README.md` for current positioning, resource count, category names, community links, badges, and disclaimer.\n- Read `resources/resources.json` or `resources/README.md` when category counts or resource types matter.\n- Read `outreach/*.md` when the task is about repurposing existing promotional copy.\n- Read `reports/availability.json` only when the article needs current availability examples or status data.\n- Read `CONTRIBUTING.md` when explaining community contribution.\n\nUse `references/brand-profile.md` for stable brand language and prohibited claims.\nUse `references/platform-rules.md` for platform-specific writing rules.\nUse `references/quality-checklist.md` before final delivery.\n\nDo not reuse outdated numbers from old outreach drafts. Recalculate or read current files when saying \"currently collected X resources\", category counts, Star counts, Telegram handle, or publication status.\n\n## Required Research\n\nUse web research whenever facts may have changed, including:\n\n- GitHub Stars, topics, trending, external collection status, or repository metadata\n- Platform rules, editorial requirements, submission URLs, or current community norms\n- Third-party tools, TVBox/影视仓 ecosystem details, IPTV project details, software versions, or policy-sensitive facts\n- Any claim using \"latest\", \"currently\", \"2026\", \"recently\", \"now\", or similar time-sensitive wording\n- User-provided URLs that need to be summarized or cited\n\nPure rewriting of supplied text can skip web research if no new facts are added.\n\nUse sources in this order:\n\n1. This repository and official project pages\n2. Official docs, project repos, platform guidelines, government/standard sources\n3. Reputable media, community posts, and high-quality technical articles\n4. User-provided materials\n\nDo not cite search result snippets, anonymous reposts, or low-quality aggregation pages as factual support.\n\n## Content Strategy\n\nEach article needs one primary intent:\n\n- **Definition**: what awesome-zhuiju-free is, who should use it\n- **Tutorial**: how to find resources, use TVBox configs, subscribe to updates, submit resources\n- **Comparison**: ordinary navigation sites vs this open-source live list\n- **Decision**: whether this route suits a specific reader\n- **List**: methods, tools, categories, resources, contribution ideas\n- **Technical**: GitHub Actions checks, JSON data, schema validation, issue-driven publishing\n- **Growth story**: how the project gained attention, what worked, what did not\n- **Community post**: ask for feedback, recruit maintainers, invite resource recommendations\n\nSet a clear center judgment before writing. The article should answer what the reader can do next, not only introduce the project.\n\n## Writing Rules\n\n### Opening\n\nWithin the first 150 Chinese characters:\n\n- Answer the core question or state the central conclusion\n- Identify who the article is for\n- Preview the practical value\n\nAvoid empty openers such as \"随着互联网的发展\", \"在当今数字化时代\", \"众所周知\", \"本文将深入探讨\", and long suspense setups.\n\n### GEO Answer Blocks\n\nInclude at least 3 self-contained answer blocks in normal article sections. Each block should:\n\n- Directly answer a likely user question\n- Name the concrete entity, such as `Awesome Zhuiju Free`, `GitHub 仓库 laoma528/awesome-zhuiju-free`, `TVBox 配置地址`\n- Include conditions, limitations, or time scope when relevant\n- Be understandable without reading the surrounding article\n- Avoid vague pronouns like \"这个\", \"它\", \"上述\"\n\nDo not label them as \"AI 引用块\".\n\n### Style\n\n- Write answer-first, then explain.\n- Use concrete nouns and actions.\n- Keep paragraphs varied and readable.\n- Use tables only when comparison or scanning is improved.\n- Avoid mechanical \"首先、其次、最后\" chains.\n- Avoid exaggerated marketing words such as \"全网最全\", \"永久免费可用\", \"神器\", \"闭眼入\", \"百分百\".\n- Do not invent personal experience, user testimonials, editor endorsements, rankings, tests, or traffic numbers.\n\n### Project Claim Boundaries\n\nAlways observe these boundaries:\n\n- The project is a curated open-source navigation/list, not a content host.\n- Do not say the project stores, distributes, cracks, or provides copyrighted video files.\n- Do not promise that third-party resources are always available, legal, safe, ad-free forever, or available in every region.\n- Say availability checks reflect the GitHub Actions environment at the check time.\n- Explain that \"free/ad-free\" is a curation criterion and best-effort manual screening, not a legal guarantee or permanent guarantee.\n- For copyright-sensitive contexts, include one concise boundary note near the relevant section or at the end.\n\n## Brand and Link Use\n\nRead `references/brand-profile.md` before adding project links.\n\nDefault link priority:\n\n1. GitHub: `https://github.com/laoma528/awesome-zhuiju-free`\n2. Website: `https://zhuiju.me`\n3. Gitee mirror: `https://gitee.com/laoma2053/awesome-zhuiju-free`\n4. Telegram channel only if current README confirms it\n\nUse links only where they help the reader. A normal article should mention the project 1-3 times, unless the whole article is a project announcement.\nWhen choosing only one primary CTA, prefer the GitHub repository unless the target platform is aimed at non-technical users who mainly need a direct browsing page.\n\nPrefer descriptive anchors:\n\n- `免费无广告追剧资源导航`\n- `查看每日检测状态`\n- `提交新资源或报告失效`\n- `TVBox/影视仓配置地址清单`\n\nAvoid naked repeated links, \"点这里\", and anchors unrelated to the target page.\n\n## Platform Adaptation\n\nRead `references/platform-rules.md` and adapt structure, not just title.\n\nCommon defaults:\n\n- **Self-hosted blog / Typecho**: complete SEO metadata, FAQ, slug, summary, internal link suggestions, update date.\n- **Zhihu**: natural answer style, weak external-link footprint, answer the question before introducing the project.\n- **Juejin / SegmentFault / CSDN / Blogyuan**: emphasize implementation, automation, data structure, reproducibility, and project operation lessons.\n- **SSPai / Appinn**: emphasize clean experience, concrete user pain, no-ad curation, and practical workflow.\n- **WeChat public account**: short paragraphs, one CTA, less SEO metadata in the body.\n- **V2EX / forums**: discussion-first, ask for feedback, be transparent about limits, avoid hard-selling.\n- **Xiaohongshu long note**: shorter, more scannable, with cover text and carousel outline when requested.\n\n## Output Format\n\nUnless the user asks for \"only the article\", deliver:\n\n### A. Writing Position\n\n- Primary intent\n- Target reader\n- Center conclusion\n- Unique value\n\n### B. Title Options\n\nProvide 3 titles and mark the recommended one.\n\n### C. Final Article\n\nProvide publishable Markdown body.\n\nWhen sources are used:\n\n- Use inline natural links if the platform supports them.\n- Use a \"参考资料\" section when links should be separated.\n- Do not include unverified or unused links.\n\n### D. Publishing Metadata\n\nInclude when appropriate for the platform:\n\n- SEO Title\n- Meta Description\n- Slug\n- Summary\n- Tags\n- Category\n- Cover text\n- Image positions and alt suggestions\n\n### E. GEO Enhancements\n\n- FAQ 3-6 items\n- Internal link suggestions\n- Optional JSON-LD type for self-hosted blog\n- Facts that should be refreshed before publishing\n\n### F. Verification Notes\n\nBriefly list:\n\n- Verification date\n- Local files and external sources used\n- Uncertainties or claims that should be avoided\n\nIf the user requests a forum post, social platform answer, or \"只要正文\", omit metadata that would not be pasted into that platform.\n\n## File Output\n\nWhen asked to save output:\n\n- Default directory: `outreach/generated/`\n- Filename: `YYYY-MM-DD-topic-slug.md`\n- Do not overwrite existing files; append `-v2`, `-v3`\n- Use UTF-8\n\n## Final Review\n\nBefore delivery, revise the draft against `references/quality-checklist.md`. Fix problems directly.\n\nThe final answer must not contain hidden reasoning, model self-evaluation, prompt residue, or claims that the article is guaranteed to rank or be cited by ChatGPT, Gemini, Perplexity, Google AI Overview, or any other system.\n","skills/geo-article-writer/references/quality-checklist.md":"# Publication Quality Checklist\n\nUse this before delivering final content. Fix issues directly.\n\n## Facts\n\n- [ ] Current resource count and category counts were read from repository files when mentioned.\n- [ ] Star counts, external collection status, Telegram handle, and platform submission status were verified when mentioned.\n- [ ] Dates use exact dates for time-sensitive facts.\n- [ ] External sources actually support the claims attached to them.\n- [ ] No invented tests, rankings, quotes, user numbers, or endorsements.\n- [ ] Availability-check claims explain that status reflects a check environment and check time.\n\n## Project Safety\n\n- [ ] The project is described as a navigation/list/status project, not a hosting or distribution service.\n- [ ] No wording implies piracy, cracking, bypassing payment, or guaranteed access.\n- [ ] \"Free/ad-free\" is expressed as a curation criterion or best-effort screening, not a permanent guarantee.\n- [ ] Copyright and third-party responsibility boundary is included where the topic is sensitive.\n- [ ] No third-party resource is promised to be legal, safe, or permanently available.\n\n## GEO Quality\n\n- [ ] The opening gives a direct answer or central conclusion.\n- [ ] There are at least 3 self-contained answer blocks.\n- [ ] Important definitions name the entity clearly and avoid vague pronouns.\n- [ ] The article covers likely follow-up questions.\n- [ ] FAQ items are real reader questions, not empty schema stuffing.\n- [ ] Unique value is present: workflow, comparison, table, checklist, example, or maintainable framework.\n\n## SEO Quality\n\n- [ ] Title matches the primary search intent.\n- [ ] H1/H2/H3 hierarchy is clean.\n- [ ] Main keyword appears naturally in title, opening, at least one heading, and conclusion when appropriate.\n- [ ] Meta Description can stand alone.\n- [ ] Slug is short, lowercase, and hyphenated for self-hosted posts.\n- [ ] Internal links use descriptive anchors.\n- [ ] External links are relevant and verified.\n\n## Naturalness\n\n- [ ] No empty opening like \"随着时代的发展\".\n- [ ] No mechanical \"首先、其次、再次、最后\" chain.\n- [ ] Section structures vary naturally.\n- [ ] No repeated transition phrases or repeated conclusions.\n- [ ] No exaggerated marketing words.\n- [ ] No fake personal experience.\n- [ ] Paragraph length supports reading on the target platform.\n\n## Platform Fit\n\n- [ ] Zhihu versions answer the question before mentioning the project.\n- [ ] Technical platform versions include implementation details and verification.\n- [ ] SSPai/Appinn versions prioritize clean user experience and practical use.\n- [ ] Forum versions invite discussion and avoid hard-selling.\n- [ ] WeChat versions use one clear CTA.\n- [ ] Xiaohongshu versions include cover/carousel guidance when needed.\n- [ ] SEO metadata is not mixed into platforms where it should not be pasted.\n\n## Final Delivery\n\n- [ ] The body can be copied directly into the target platform.\n- [ ] Metadata, title, tags, and summary match the article.\n- [ ] Markdown tables, links, and code blocks render correctly.\n- [ ] No internal analysis, prompt text, or model self-review remains.\n- [ ] No claim says the article is guaranteed to rank or be cited by AI search.\n","CLAUDE.md":"# CLAUDE.md\n\nThis file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.\n\n## 项目概述\n\n这是一个面向中文用户的免费影视资源导航列表（awesome-list 风格）。核心数据存储在 `resources/resources.json`，README.md 由脚本从数据自动生成，不应手动编辑 README 中的精选资源区块。\n\n## 常用命令\n\n所有脚本均为原生 Node.js ESM 模块，**无需 `npm install`**，直接运行：\n\n```bash\n# 验证所有 JSON 数据格式合法性\nnode scripts/validate-data.mjs\n\n# 将 resources.json 中的精选资源同步到 README.md（重建表格）\nnode scripts/sync-readme.mjs\n\n# 检测所有精选资源的 URL 可访问性，更新 reports/availability.json 和 README\nnode scripts/check-availability.mjs\n\n# 更新 open_source 类型资源的 GitHub star 数和最近推送时间\nGITHUB_TOKEN=<token> node scripts/update-open-source-stats.mjs\n```\n\nCI 使用 Node.js 24；本地需 Node.js 18+（使用了原生 `fetch` 和顶层 `await`）。\n\n## 架构与数据流\n\n### 单一数据源\n\n`resources/resources.json` 是所有资源的唯一数据源。README 中 `<!-- featured-resources:start -->` 和 `<!-- featured-resources:end -->` 之间的全部内容，以及 `<!-- resource-count:start/end -->` 徽章，**均由 `scripts/sync-readme.mjs` 自动覆盖**，手动编辑这些区块会在下次 CI 运行时被还原。\n\n### 脚本职责\n\n| 脚本 | 输入 | 输出 |\n|---|---|---|\n| `sync-readme.mjs` | `resources/resources.json`，`reports/availability.json` | `README.md`（精选区块） |\n| `check-availability.mjs` | `resources/resources.json`，`README.md` | `reports/availability.json`，`README.md`（状态标记） |\n| `validate-data.mjs` | 全部 JSON 文件 | 仅 stdout 错误报告 |\n| `update-open-source-stats.mjs` | `resources/resources.json`（GitHub API） | `resources/resources.json` |\n\n### 自动化流程\n\n- **每日 09:00 (北京时间)**：`check-availability.yml` 依次执行 sync-readme → check-availability → validate-data，提交结果\n- **PR / push to main**：`validate-data.yml` 校验数据格式\n- **Issue 审核**：维护者在资源推荐 Issue 下评论 `ok` 触发 `issue-to-resource-pr.yml`，自动将资源写入 resources.json 并发布\n\n## resources.json 数据约束\n\n新增或修改资源时必须满足以下规则（`validate-data.mjs` 会检测）：\n\n- `id`：小写字母和数字，以 `-` 分隔，模式 `^[a-z0-9]+(?:-[a-z0-9]+)*$`，全局唯一\n- `url`：必须是 HTTPS，`tvbox_config` 类型例外可用 HTTP\n- `category`：必须是枚举值之一：`online_video` / `video_app` / `cloud_search` / `magnet_search` / `subtitles` / `player` / `subscription` / `tvbox_config` / `membership` / `open_source` / `other`\n- `scores.*`：`more` / `speed` / `clean` / `stable` / `ease` 均为 1.0–5.0，最多一位小数\n- `risks.*`：`copyright` / `safety` / `privacy` / `payment` 枚举值：`low` / `medium` / `high` / `unknown`\n- `verification.status`：`pending` / `verified` / `recommended` / `caution` / `temporarily_unavailable` / `removed`\n- `featured: true` 的资源必须提供 `summary_short`（40 个字符以内）\n- `category: \"open_source\"` 的资源必须包含 `github.full_name`（`owner/repo` 格式）、`github.stars`（非负整数）、`github.pushed_at`（ISO 时间戳）\n- 日期字段格式：`YYYY-MM-DD`\n\n修改 resources.json 后，手动运行 `node scripts/sync-readme.mjs` 可在本地预览 README 变化。\n","skills/geo-article-writer/agents/openai.yaml":"interface:\n  display_name: \"GEO Article Writer\"\n  short_description: \"为 awesome-zhuiju-free 生成 GitHub 优先的 GEO 宣传长文\"\n  default_prompt: \"Use $geo-article-writer to write a GitHub-first promotional article for awesome-zhuiju-free.\"\n\npolicy:\n  allow_implicit_invocation: true\n","skills/geo-article-writer/references/platform-rules.md":"# Platform Rules\n\nAdapt the same topic to the target platform. Do not merely change the title.\n\n## Self-hosted Blog / Typecho\n\nGoal: search traffic plus AI-search-friendly answers.\n\n- Use one H1 title, H2/H3 body.\n- Put the direct answer in the first 150 Chinese characters.\n- Include 3+ answer-first sections.\n- Add FAQ when it reflects real reader questions.\n- Include SEO Title, Meta Description, Slug, summary, tags, category, and image alt suggestions.\n- Suggest Article or BlogPosting JSON-LD; suggest FAQPage only when FAQ is substantial.\n- Include \"最后核验：YYYY-MM-DD\" for time-sensitive counts or platform facts.\n- Add 1-3 natural links, with the GitHub repository as the default primary link and zhuiju.me as a secondary browsing link.\n\n## Zhihu\n\nGoal: answer the question convincingly without looking like an ad.\n\n- Start with a direct judgment or practical answer.\n- Mention the project only after giving standalone value.\n- Usually use at most one external link in the answer body; use the GitHub repository by default unless the answer is purely for ordinary users who need direct browsing.\n- Avoid SEO metadata, Schema, and promotional slogans.\n- Use fewer headings than a blog article.\n- Include tradeoffs, limitations, and alternative methods.\n- Do not ask for Star unless the question itself is about open-source support.\n\n## Juejin / SegmentFault\n\nGoal: technical story, implementation logic, or growth review.\n\n- Prefer topics such as GitHub Actions availability checks, structured JSON data, schema validation, issue-driven workflow, README conversion, and open-source growth.\n- Explain the problem, design choices, implementation outline, and tradeoffs.\n- Include code snippets only when they are accurate and useful.\n- Mark example code as simplified if not copied from repository.\n- Include what did not work or what remains limited when writing a growth review.\n- End with a useful principle or next step, not just a project link.\n\n## CSDN / 51CTO / Blogyuan\n\nGoal: reproducible technical documentation.\n\n- State environment, files, workflow, and expected result.\n- Keep commands copyable and explain where to run them.\n- Include verification and rollback/repair advice when editing workflows.\n- For GitHub Actions or platform behavior, verify current docs if version-sensitive.\n- Use troubleshooting tables for failures.\n- Avoid filler background and non-technical promotion.\n\n## SSPai Matrix\n\nGoal: practical tool article for readers who value clean experience.\n\n- Start from a real everyday pain point.\n- Explain the workflow: open page, scan status, choose category, submit issue if needed.\n- Emphasize clean, open, transparent, and maintained.\n- Keep technical explanation short and readable.\n- Use screenshots: README first screen, resource table with status, website browsing view.\n- End with a concise boundary note, not aggressive CTA.\n\n## Appinn / 小众软件\n\nGoal: concise discovery post for clean tools.\n\n- Format like a software recommendation.\n- Lead with labels such as `免费`、`开源`、`Web`.\n- Focus on no-ad curation, daily checks, and direct use without registration.\n- Keep language restrained; do not ask for Star.\n- Include platform, price, website, GitHub, and tags if needed.\n\n## V2EX / NodeSeek / HostLoc / Forums\n\nGoal: discussion and feedback, not one-way advertising.\n\n- Use first-person project context only if true or supplied by the user.\n- Ask concrete questions: categories to add, README clarity, TVBox configs, maintainers.\n- Be transparent about copyright and availability limitations.\n- Avoid \"求 Star\", \"全网最全\", and repeated links.\n- Keep the tone direct and community-native.\n\n## WeChat Public Account\n\nGoal: readable article or editor-ready contribution.\n\n- Use short paragraphs and clear section headings.\n- Keep one CTA. Prefer GitHub Star, GitHub repository visit, or GitHub Issue contribution; use website browsing CTA only for non-technical resource/tool accounts.\n- Put references at the end if needed.\n- Avoid dense SEO keyword repetition.\n- For GitHub-oriented accounts, include the technical differentiation and growth angle.\n- For resource/tool accounts, focus on the user pain and daily availability check.\n\n## Xiaohongshu Long Note\n\nGoal: scannable note with save/share value.\n\n- Keep the main text around 600-1200 Chinese characters unless requested otherwise.\n- First 100 characters must state who it helps and what problem it solves.\n- Use practical labels: 免费追剧、无广告、TVBox配置、电视盒子、资源失效.\n- Provide cover text and 3-6 page carousel outline when requested.\n- Avoid dense external links; direct readers to search the project name or the GitHub repository name if link placement is limited. Mention zhuiju.me only as the direct browsing entry.\n\n## Blogger / Editor Pitch\n\nGoal: make forwarding or coverage easy.\n\n- Provide a short personal note plus a ready-to-publish blurb.\n- Include 3 facts: what it is, why it is different, where to see it.\n- Keep the ask lightweight.\n- Do not overstate Star growth, media coverage, or uniqueness.\n","skills/geo-article-writer/references/brand-profile.md":"# Brand Profile: awesome-zhuiju-free\n\nUse this file to keep article claims and link placement aligned with the project.\n\n## Default Brand\n\n### Awesome Zhuiju Free\n\n- Chinese name/description: 免费无广告追剧资源导航\n- Primary URL: `https://github.com/laoma528/awesome-zhuiju-free`\n- Secondary browsing URL: `https://zhuiju.me`\n- Gitee mirror: `https://gitee.com/laoma2053/awesome-zhuiju-free`\n- Positioning: A curated open-source list and website for Chinese streaming-related resources, with daily automated availability checks.\n- Core promise: Help users quickly see which curated resources are currently reachable, instead of manually opening many dead or ad-heavy links.\n\n## Recommended Wording\n\nUse these phrases naturally:\n\n- `Awesome Zhuiju Free`\n- `GitHub 仓库 laoma528/awesome-zhuiju-free`\n- `zhuiju.me`\n- `免费无广告追剧资源导航`\n- `开源追剧资源清单`\n- `每日自动检测可用性`\n- `人工精选、社区共同维护`\n- `TVBox / 影视仓配置地址清单`\n- `在线影视、影视APP、网盘搜索、磁力 BT、字幕、IPTV、开源播放器项目`\n\n## Stable Selling Points\n\nUse only when supported by current repository files:\n\n- Open-source list with public structured data.\n- Curated resources are screened for free and ad-free experience as a project criterion.\n- GitHub Actions runs availability checks and writes status data.\n- README and website show status such as reachable, restricted, unavailable, or unchecked.\n- Community can recommend resources or report failures through GitHub Issues.\n- Data structure supports schema validation and repeatable maintenance.\n\nWhen using counts, read current files first. Do not copy stale outreach numbers.\n\n## Link Priority\n\nUse links according to reader context:\n\n1. `https://github.com/laoma528/awesome-zhuiju-free` as the default primary link for most articles, pitches, answers, and open-source promotion.\n2. `https://zhuiju.me` as the secondary browsing link for ordinary users who want to view the resource list directly.\n3. `https://gitee.com/laoma2053/awesome-zhuiju-free` when discussing domestic access or GitHub access issues.\n4. Telegram or other community links only after confirming the current README.\n\nIf a platform or article allows only one link, use the GitHub repository by default. Use `zhuiju.me` as the only link only for strongly non-technical contexts where direct browsing matters more than GitHub Star or contribution conversion.\n\n## CTA Patterns\n\nChoose one CTA, not several:\n\n- Star support: `如果这个清单对你有帮助，可以在 GitHub 点一个 Star，方便后续找到，也能帮助更多人发现这个项目。`\n- Repository visit: `完整资源清单、每日检测状态和贡献入口都在 GitHub 仓库中维护。`\n- Browse resources: `普通用户也可以打开 zhuiju.me 直接查看今天可用的资源状态。`\n- Contribute: `发现好用的免费无广告资源，可以通过 GitHub Issue 推荐。`\n- Report failure: `如果某个资源失效，可以提交 Issue 帮助维护列表。`\n- Subscribe: `订阅频道获取资源更新通知。` Use only if the README confirms a channel.\n\n## Prohibited Claims\n\nDo not claim or imply:\n\n- The project hosts, stores, distributes, cracks, unlocks, or provides copyrighted video files.\n- All resources are legal, permanently valid, permanently safe, or available in every region.\n- Availability status is a universal guarantee. It only reflects the check environment and check time.\n- \"全网最全\", \"永久免费\", \"百分百无广告\", \"一定能看\", \"官方片源\", \"免会员破解\".\n- Any Star number, ranking, media recommendation, collection status, or user scale without current verification.\n- Paid promotion neutrality beyond what is documented. If saying \"不卖推广位\", verify it from current project text or user-provided instructions.\n\n## Compliance Boundary\n\nFor copyright-sensitive articles, include one concise note:\n\n> Awesome Zhuiju Free 是资源信息导航和可用性状态整理项目，不存储或传播影视文件。第三方资源的内容、访问规则和合规性以对应站点为准。\n\nDo not repeat the note in every section.\n\n## Tone by Audience\n\n- Ordinary streaming users: practical, low-jargon, focus on saving time and avoiding dead links.\n- TVBox/影视仓 users: focus on config address stability, multi-source backup, and update checks.\n- Open-source readers: focus on data transparency, GitHub Actions, schema validation, contribution workflow.\n- Editors/bloggers: focus on why the project is useful, credible, and easy to introduce.\n- Forum readers: transparent and discussion-oriented; ask for feedback and resource recommendations.\n"},"files":{"skills/geo-article-writer/SKILL.md":"---\nname: geo-article-writer\ndescription: Write, rewrite, adapt, or review Chinese GEO/SEO long-form content that primarily promotes the GitHub repository laoma528/awesome-zhuiju-free, with zhuiju.me as a secondary browsing entry. Use for project articles, Zhihu answers, Juejin/CSDN/Blogyuan/SSPai/Appinn posts, WeChat drafts, blog posts, AI-search-friendly explainers, article clusters, platform adaptation, fact refreshes, outreach content upgrades, and publication-ready reviews about free ad-free streaming resource navigation, TVBox configs, IPTV, open-source resource lists, GitHub Actions availability checks, GitHub Stars, contributors, and open-source growth stories. Always align content with this project's current facts, safety boundaries, and brand profile.\n---\n\n# GEO Article Writer\n\nProduce publishable Chinese long-form content for `awesome-zhuiju-free`: topic research, article structure, full draft, GEO/SEO optimization, platform adaptation, and final review.\n\nThis skill is not a generic marketing writer. It should turn the project's real assets into useful articles that can be cited by search engines, AI search tools, community readers, and platform editors without overstating what the project provides.\n\n## Core Inputs\n\nExtract these from the user request:\n\n- Topic, keyword, draft, target question, or target platform\n- Desired mode: create, rewrite, review, platform-adapt, cluster, update\n- Target reader: ordinary streaming user, TVBox/影视仓 user, GitHub/open-source reader, technical operator, editor/blogger\n- Goal: search traffic, AI search visibility, community discussion, project introduction, contributor recruitment, or external submission\n- Required facts, links, screenshots, or existing outreach files\n- Forbidden claims, sensitive wording, length, tone, and publication date\n\nIf the user gives only a topic, default to `create`.\nIf the user gives a draft, default to `rewrite`.\nIf the user asks whether an article is good enough, default to `review` and deliver a revised final version.\nIf the user names a platform, adapt to that platform instead of producing a generic blog article.\n\nDefault output language is Simplified Chinese.\n\n## Project Context\n\nBefore writing project-related claims, prefer local project files over memory:\n\n- Read `README.md` for current positioning, resource count, category names, community links, badges, and disclaimer.\n- Read `resources/resources.json` or `resources/README.md` when category counts or resource types matter.\n- Read `outreach/*.md` when the task is about repurposing existing promotional copy.\n- Read `reports/availability.json` only when the article needs current availability examples or status data.\n- Read `CONTRIBUTING.md` when explaining community contribution.\n\nUse `references/brand-profile.md` for stable brand language and prohibited claims.\nUse `references/platform-rules.md` for platform-specific writing rules.\nUse `references/quality-checklist.md` before final delivery.\n\nDo not reuse outdated numbers from old outreach drafts. Recalculate or read current files when saying \"currently collected X resources\", category counts, Star counts, Telegram handle, or publication status.\n\n## Required Research\n\nUse web research whenever facts may have changed, including:\n\n- GitHub Stars, topics, trending, external collection status, or repository metadata\n- Platform rules, editorial requirements, submission URLs, or current community norms\n- Third-party tools, TVBox/影视仓 ecosystem details, IPTV project details, software versions, or policy-sensitive facts\n- Any claim using \"latest\", \"currently\", \"2026\", \"recently\", \"now\", or similar time-sensitive wording\n- User-provided URLs that need to be summarized or cited\n\nPure rewriting of supplied text can skip web research if no new facts are added.\n\nUse sources in this order:\n\n1. This repository and official project pages\n2. Official docs, project repos, platform guidelines, government/standard sources\n3. Reputable media, community posts, and high-quality technical articles\n4. User-provided materials\n\nDo not cite search result snippets, anonymous reposts, or low-quality aggregation pages as factual support.\n\n## Content Strategy\n\nEach article needs one primary intent:\n\n- **Definition**: what awesome-zhuiju-free is, who should use it\n- **Tutorial**: how to find resources, use TVBox configs, subscribe to updates, submit resources\n- **Comparison**: ordinary navigation sites vs this open-source live list\n- **Decision**: whether this route suits a specific reader\n- **List**: methods, tools, categories, resources, contribution ideas\n- **Technical**: GitHub Actions checks, JSON data, schema validation, issue-driven publishing\n- **Growth story**: how the project gained attention, what worked, what did not\n- **Community post**: ask for feedback, recruit maintainers, invite resource recommendations\n\nSet a clear center judgment before writing. The article should answer what the reader can do next, not only introduce the project.\n\n## Writing Rules\n\n### Opening\n\nWithin the first 150 Chinese characters:\n\n- Answer the core question or state the central conclusion\n- Identify who the article is for\n- Preview the practical value\n\nAvoid empty openers such as \"随着互联网的发展\", \"在当今数字化时代\", \"众所周知\", \"本文将深入探讨\", and long suspense setups.\n\n### GEO Answer Blocks\n\nInclude at least 3 self-contained answer blocks in normal article sections. Each block should:\n\n- Directly answer a likely user question\n- Name the concrete entity, such as `Awesome Zhuiju Free`, `GitHub 仓库 laoma528/awesome-zhuiju-free`, `TVBox 配置地址`\n- Include conditions, limitations, or time scope when relevant\n- Be understandable without reading the surrounding article\n- Avoid vague pronouns like \"这个\", \"它\", \"上述\"\n\nDo not label them as \"AI 引用块\".\n\n### Style\n\n- Write answer-first, then explain.\n- Use concrete nouns and actions.\n- Keep paragraphs varied and readable.\n- Use tables only when comparison or scanning is improved.\n- Avoid mechanical \"首先、其次、最后\" chains.\n- Avoid exaggerated marketing words such as \"全网最全\", \"永久免费可用\", \"神器\", \"闭眼入\", \"百分百\".\n- Do not invent personal experience, user testimonials, editor endorsements, rankings, tests, or traffic numbers.\n\n### Project Claim Boundaries\n\nAlways observe these boundaries:\n\n- The project is a curated open-source navigation/list, not a content host.\n- Do not say the project stores, distributes, cracks, or provides copyrighted video files.\n- Do not promise that third-party resources are always available, legal, safe, ad-free forever, or available in every region.\n- Say availability checks reflect the GitHub Actions environment at the check time.\n- Explain that \"free/ad-free\" is a curation criterion and best-effort manual screening, not a legal guarantee or permanent guarantee.\n- For copyright-sensitive contexts, include one concise boundary note near the relevant section or at the end.\n\n## Brand and Link Use\n\nRead `references/brand-profile.md` before adding project links.\n\nDefault link priority:\n\n1. GitHub: `https://github.com/laoma528/awesome-zhuiju-free`\n2. Website: `https://zhuiju.me`\n3. Gitee mirror: `https://gitee.com/laoma2053/awesome-zhuiju-free`\n4. Telegram channel only if current README confirms it\n\nUse links only where they help the reader. A normal article should mention the project 1-3 times, unless the whole article is a project announcement.\nWhen choosing only one primary CTA, prefer the GitHub repository unless the target platform is aimed at non-technical users who mainly need a direct browsing page.\n\nPrefer descriptive anchors:\n\n- `免费无广告追剧资源导航`\n- `查看每日检测状态`\n- `提交新资源或报告失效`\n- `TVBox/影视仓配置地址清单`\n\nAvoid naked repeated links, \"点这里\", and anchors unrelated to the target page.\n\n## Platform Adaptation\n\nRead `references/platform-rules.md` and adapt structure, not just title.\n\nCommon defaults:\n\n- **Self-hosted blog / Typecho**: complete SEO metadata, FAQ, slug, summary, internal link suggestions, update date.\n- **Zhihu**: natural answer style, weak external-link footprint, answer the question before introducing the project.\n- **Juejin / SegmentFault / CSDN / Blogyuan**: emphasize implementation, automation, data structure, reproducibility, and project operation lessons.\n- **SSPai / Appinn**: emphasize clean experience, concrete user pain, no-ad curation, and practical workflow.\n- **WeChat public account**: short paragraphs, one CTA, less SEO metadata in the body.\n- **V2EX / forums**: discussion-first, ask for feedback, be transparent about limits, avoid hard-selling.\n- **Xiaohongshu long note**: shorter, more scannable, with cover text and carousel outline when requested.\n\n## Output Format\n\nUnless the user asks for \"only the article\", deliver:\n\n### A. Writing Position\n\n- Primary intent\n- Target reader\n- Center conclusion\n- Unique value\n\n### B. Title Options\n\nProvide 3 titles and mark the recommended one.\n\n### C. Final Article\n\nProvide publishable Markdown body.\n\nWhen sources are used:\n\n- Use inline natural links if the platform supports them.\n- Use a \"参考资料\" section when links should be separated.\n- Do not include unverified or unused links.\n\n### D. Publishing Metadata\n\nInclude when appropriate for the platform:\n\n- SEO Title\n- Meta Description\n- Slug\n- Summary\n- Tags\n- Category\n- Cover text\n- Image positions and alt suggestions\n\n### E. GEO Enhancements\n\n- FAQ 3-6 items\n- Internal link suggestions\n- Optional JSON-LD type for self-hosted blog\n- Facts that should be refreshed before publishing\n\n### F. Verification Notes\n\nBriefly list:\n\n- Verification date\n- Local files and external sources used\n- Uncertainties or claims that should be avoided\n\nIf the user requests a forum post, social platform answer, or \"只要正文\", omit metadata that would not be pasted into that platform.\n\n## File Output\n\nWhen asked to save output:\n\n- Default directory: `outreach/generated/`\n- Filename: `YYYY-MM-DD-topic-slug.md`\n- Do not overwrite existing files; append `-v2`, `-v3`\n- Use UTF-8\n\n## Final Review\n\nBefore delivery, revise the draft against `references/quality-checklist.md`. Fix problems directly.\n\nThe final answer must not contain hidden reasoning, model self-evaluation, prompt residue, or claims that the article is guaranteed to rank or be cited by ChatGPT, Gemini, Perplexity, Google AI Overview, or any other system.\n","skills/geo-article-writer/references/quality-checklist.md":"# Publication Quality Checklist\n\nUse this before delivering final content. Fix issues directly.\n\n## Facts\n\n- [ ] Current resource count and category counts were read from repository files when mentioned.\n- [ ] Star counts, external collection status, Telegram handle, and platform submission status were verified when mentioned.\n- [ ] Dates use exact dates for time-sensitive facts.\n- [ ] External sources actually support the claims attached to them.\n- [ ] No invented tests, rankings, quotes, user numbers, or endorsements.\n- [ ] Availability-check claims explain that status reflects a check environment and check time.\n\n## Project Safety\n\n- [ ] The project is described as a navigation/list/status project, not a hosting or distribution service.\n- [ ] No wording implies piracy, cracking, bypassing payment, or guaranteed access.\n- [ ] \"Free/ad-free\" is expressed as a curation criterion or best-effort screening, not a permanent guarantee.\n- [ ] Copyright and third-party responsibility boundary is included where the topic is sensitive.\n- [ ] No third-party resource is promised to be legal, safe, or permanently available.\n\n## GEO Quality\n\n- [ ] The opening gives a direct answer or central conclusion.\n- [ ] There are at least 3 self-contained answer blocks.\n- [ ] Important definitions name the entity clearly and avoid vague pronouns.\n- [ ] The article covers likely follow-up questions.\n- [ ] FAQ items are real reader questions, not empty schema stuffing.\n- [ ] Unique value is present: workflow, comparison, table, checklist, example, or maintainable framework.\n\n## SEO Quality\n\n- [ ] Title matches the primary search intent.\n- [ ] H1/H2/H3 hierarchy is clean.\n- [ ] Main keyword appears naturally in title, opening, at least one heading, and conclusion when appropriate.\n- [ ] Meta Description can stand alone.\n- [ ] Slug is short, lowercase, and hyphenated for self-hosted posts.\n- [ ] Internal links use descriptive anchors.\n- [ ] External links are relevant and verified.\n\n## Naturalness\n\n- [ ] No empty opening like \"随着时代的发展\".\n- [ ] No mechanical \"首先、其次、再次、最后\" chain.\n- [ ] Section structures vary naturally.\n- [ ] No repeated transition phrases or repeated conclusions.\n- [ ] No exaggerated marketing words.\n- [ ] No fake personal experience.\n- [ ] Paragraph length supports reading on the target platform.\n\n## Platform Fit\n\n- [ ] Zhihu versions answer the question before mentioning the project.\n- [ ] Technical platform versions include implementation details and verification.\n- [ ] SSPai/Appinn versions prioritize clean user experience and practical use.\n- [ ] Forum versions invite discussion and avoid hard-selling.\n- [ ] WeChat versions use one clear CTA.\n- [ ] Xiaohongshu versions include cover/carousel guidance when needed.\n- [ ] SEO metadata is not mixed into platforms where it should not be pasted.\n\n## Final Delivery\n\n- [ ] The body can be copied directly into the target platform.\n- [ ] Metadata, title, tags, and summary match the article.\n- [ ] Markdown tables, links, and code blocks render correctly.\n- [ ] No internal analysis, prompt text, or model self-review remains.\n- [ ] No claim says the article is guaranteed to rank or be cited by AI search.\n","CLAUDE.md":"# CLAUDE.md\n\nThis file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.\n\n## 项目概述\n\n这是一个面向中文用户的免费影视资源导航列表（awesome-list 风格）。核心数据存储在 `resources/resources.json`，README.md 由脚本从数据自动生成，不应手动编辑 README 中的精选资源区块。\n\n## 常用命令\n\n所有脚本均为原生 Node.js ESM 模块，**无需 `npm install`**，直接运行：\n\n```bash\n# 验证所有 JSON 数据格式合法性\nnode scripts/validate-data.mjs\n\n# 将 resources.json 中的精选资源同步到 README.md（重建表格）\nnode scripts/sync-readme.mjs\n\n# 检测所有精选资源的 URL 可访问性，更新 reports/availability.json 和 README\nnode scripts/check-availability.mjs\n\n# 更新 open_source 类型资源的 GitHub star 数和最近推送时间\nGITHUB_TOKEN=<token> node scripts/update-open-source-stats.mjs\n```\n\nCI 使用 Node.js 24；本地需 Node.js 18+（使用了原生 `fetch` 和顶层 `await`）。\n\n## 架构与数据流\n\n### 单一数据源\n\n`resources/resources.json` 是所有资源的唯一数据源。README 中 `<!-- featured-resources:start -->` 和 `<!-- featured-resources:end -->` 之间的全部内容，以及 `<!-- resource-count:start/end -->` 徽章，**均由 `scripts/sync-readme.mjs` 自动覆盖**，手动编辑这些区块会在下次 CI 运行时被还原。\n\n### 脚本职责\n\n| 脚本 | 输入 | 输出 |\n|---|---|---|\n| `sync-readme.mjs` | `resources/resources.json`，`reports/availability.json` | `README.md`（精选区块） |\n| `check-availability.mjs` | `resources/resources.json`，`README.md` | `reports/availability.json`，`README.md`（状态标记） |\n| `validate-data.mjs` | 全部 JSON 文件 | 仅 stdout 错误报告 |\n| `update-open-source-stats.mjs` | `resources/resources.json`（GitHub API） | `resources/resources.json` |\n\n### 自动化流程\n\n- **每日 09:00 (北京时间)**：`check-availability.yml` 依次执行 sync-readme → check-availability → validate-data，提交结果\n- **PR / push to main**：`validate-data.yml` 校验数据格式\n- **Issue 审核**：维护者在资源推荐 Issue 下评论 `ok` 触发 `issue-to-resource-pr.yml`，自动将资源写入 resources.json 并发布\n\n## resources.json 数据约束\n\n新增或修改资源时必须满足以下规则（`validate-data.mjs` 会检测）：\n\n- `id`：小写字母和数字，以 `-` 分隔，模式 `^[a-z0-9]+(?:-[a-z0-9]+)*$`，全局唯一\n- `url`：必须是 HTTPS，`tvbox_config` 类型例外可用 HTTP\n- `category`：必须是枚举值之一：`online_video` / `video_app` / `cloud_search` / `magnet_search` / `subtitles` / `player` / `subscription` / `tvbox_config` / `membership` / `open_source` / `other`\n- `scores.*`：`more` / `speed` / `clean` / `stable` / `ease` 均为 1.0–5.0，最多一位小数\n- `risks.*`：`copyright` / `safety` / `privacy` / `payment` 枚举值：`low` / `medium` / `high` / `unknown`\n- `verification.status`：`pending` / `verified` / `recommended` / `caution` / `temporarily_unavailable` / `removed`\n- `featured: true` 的资源必须提供 `summary_short`（40 个字符以内）\n- `category: \"open_source\"` 的资源必须包含 `github.full_name`（`owner/repo` 格式）、`github.stars`（非负整数）、`github.pushed_at`（ISO 时间戳）\n- 日期字段格式：`YYYY-MM-DD`\n\n修改 resources.json 后，手动运行 `node scripts/sync-readme.mjs` 可在本地预览 README 变化。\n","skills/geo-article-writer/agents/openai.yaml":"interface:\n  display_name: \"GEO Article Writer\"\n  short_description: \"为 awesome-zhuiju-free 生成 GitHub 优先的 GEO 宣传长文\"\n  default_prompt: \"Use $geo-article-writer to write a GitHub-first promotional article for awesome-zhuiju-free.\"\n\npolicy:\n  allow_implicit_invocation: true\n","skills/geo-article-writer/references/platform-rules.md":"# Platform Rules\n\nAdapt the same topic to the target platform. Do not merely change the title.\n\n## Self-hosted Blog / Typecho\n\nGoal: search traffic plus AI-search-friendly answers.\n\n- Use one H1 title, H2/H3 body.\n- Put the direct answer in the first 150 Chinese characters.\n- Include 3+ answer-first sections.\n- Add FAQ when it reflects real reader questions.\n- Include SEO Title, Meta Description, Slug, summary, tags, category, and image alt suggestions.\n- Suggest Article or BlogPosting JSON-LD; suggest FAQPage only when FAQ is substantial.\n- Include \"最后核验：YYYY-MM-DD\" for time-sensitive counts or platform facts.\n- Add 1-3 natural links, with the GitHub repository as the default primary link and zhuiju.me as a secondary browsing link.\n\n## Zhihu\n\nGoal: answer the question convincingly without looking like an ad.\n\n- Start with a direct judgment or practical answer.\n- Mention the project only after giving standalone value.\n- Usually use at most one external link in the answer body; use the GitHub repository by default unless the answer is purely for ordinary users who need direct browsing.\n- Avoid SEO metadata, Schema, and promotional slogans.\n- Use fewer headings than a blog article.\n- Include tradeoffs, limitations, and alternative methods.\n- Do not ask for Star unless the question itself is about open-source support.\n\n## Juejin / SegmentFault\n\nGoal: technical story, implementation logic, or growth review.\n\n- Prefer topics such as GitHub Actions availability checks, structured JSON data, schema validation, issue-driven workflow, README conversion, and open-source growth.\n- Explain the problem, design choices, implementation outline, and tradeoffs.\n- Include code snippets only when they are accurate and useful.\n- Mark example code as simplified if not copied from repository.\n- Include what did not work or what remains limited when writing a growth review.\n- End with a useful principle or next step, not just a project link.\n\n## CSDN / 51CTO / Blogyuan\n\nGoal: reproducible technical documentation.\n\n- State environment, files, workflow, and expected result.\n- Keep commands copyable and explain where to run them.\n- Include verification and rollback/repair advice when editing workflows.\n- For GitHub Actions or platform behavior, verify current docs if version-sensitive.\n- Use troubleshooting tables for failures.\n- Avoid filler background and non-technical promotion.\n\n## SSPai Matrix\n\nGoal: practical tool article for readers who value clean experience.\n\n- Start from a real everyday pain point.\n- Explain the workflow: open page, scan status, choose category, submit issue if needed.\n- Emphasize clean, open, transparent, and maintained.\n- Keep technical explanation short and readable.\n- Use screenshots: README first screen, resource table with status, website browsing view.\n- End with a concise boundary note, not aggressive CTA.\n\n## Appinn / 小众软件\n\nGoal: concise discovery post for clean tools.\n\n- Format like a software recommendation.\n- Lead with labels such as `免费`、`开源`、`Web`.\n- Focus on no-ad curation, daily checks, and direct use without registration.\n- Keep language restrained; do not ask for Star.\n- Include platform, price, website, GitHub, and tags if needed.\n\n## V2EX / NodeSeek / HostLoc / Forums\n\nGoal: discussion and feedback, not one-way advertising.\n\n- Use first-person project context only if true or supplied by the user.\n- Ask concrete questions: categories to add, README clarity, TVBox configs, maintainers.\n- Be transparent about copyright and availability limitations.\n- Avoid \"求 Star\", \"全网最全\", and repeated links.\n- Keep the tone direct and community-native.\n\n## WeChat Public Account\n\nGoal: readable article or editor-ready contribution.\n\n- Use short paragraphs and clear section headings.\n- Keep one CTA. Prefer GitHub Star, GitHub repository visit, or GitHub Issue contribution; use website browsing CTA only for non-technical resource/tool accounts.\n- Put references at the end if needed.\n- Avoid dense SEO keyword repetition.\n- For GitHub-oriented accounts, include the technical differentiation and growth angle.\n- For resource/tool accounts, focus on the user pain and daily availability check.\n\n## Xiaohongshu Long Note\n\nGoal: scannable note with save/share value.\n\n- Keep the main text around 600-1200 Chinese characters unless requested otherwise.\n- First 100 characters must state who it helps and what problem it solves.\n- Use practical labels: 免费追剧、无广告、TVBox配置、电视盒子、资源失效.\n- Provide cover text and 3-6 page carousel outline when requested.\n- Avoid dense external links; direct readers to search the project name or the GitHub repository name if link placement is limited. Mention zhuiju.me only as the direct browsing entry.\n\n## Blogger / Editor Pitch\n\nGoal: make forwarding or coverage easy.\n\n- Provide a short personal note plus a ready-to-publish blurb.\n- Include 3 facts: what it is, why it is different, where to see it.\n- Keep the ask lightweight.\n- Do not overstate Star growth, media coverage, or uniqueness.\n","skills/geo-article-writer/references/brand-profile.md":"# Brand Profile: awesome-zhuiju-free\n\nUse this file to keep article claims and link placement aligned with the project.\n\n## Default Brand\n\n### Awesome Zhuiju Free\n\n- Chinese name/description: 免费无广告追剧资源导航\n- Primary URL: `https://github.com/laoma528/awesome-zhuiju-free`\n- Secondary browsing URL: `https://zhuiju.me`\n- Gitee mirror: `https://gitee.com/laoma2053/awesome-zhuiju-free`\n- Positioning: A curated open-source list and website for Chinese streaming-related resources, with daily automated availability checks.\n- Core promise: Help users quickly see which curated resources are currently reachable, instead of manually opening many dead or ad-heavy links.\n\n## Recommended Wording\n\nUse these phrases naturally:\n\n- `Awesome Zhuiju Free`\n- `GitHub 仓库 laoma528/awesome-zhuiju-free`\n- `zhuiju.me`\n- `免费无广告追剧资源导航`\n- `开源追剧资源清单`\n- `每日自动检测可用性`\n- `人工精选、社区共同维护`\n- `TVBox / 影视仓配置地址清单`\n- `在线影视、影视APP、网盘搜索、磁力 BT、字幕、IPTV、开源播放器项目`\n\n## Stable Selling Points\n\nUse only when supported by current repository files:\n\n- Open-source list with public structured data.\n- Curated resources are screened for free and ad-free experience as a project criterion.\n- GitHub Actions runs availability checks and writes status data.\n- README and website show status such as reachable, restricted, unavailable, or unchecked.\n- Community can recommend resources or report failures through GitHub Issues.\n- Data structure supports schema validation and repeatable maintenance.\n\nWhen using counts, read current files first. Do not copy stale outreach numbers.\n\n## Link Priority\n\nUse links according to reader context:\n\n1. `https://github.com/laoma528/awesome-zhuiju-free` as the default primary link for most articles, pitches, answers, and open-source promotion.\n2. `https://zhuiju.me` as the secondary browsing link for ordinary users who want to view the resource list directly.\n3. `https://gitee.com/laoma2053/awesome-zhuiju-free` when discussing domestic access or GitHub access issues.\n4. Telegram or other community links only after confirming the current README.\n\nIf a platform or article allows only one link, use the GitHub repository by default. Use `zhuiju.me` as the only link only for strongly non-technical contexts where direct browsing matters more than GitHub Star or contribution conversion.\n\n## CTA Patterns\n\nChoose one CTA, not several:\n\n- Star support: `如果这个清单对你有帮助，可以在 GitHub 点一个 Star，方便后续找到，也能帮助更多人发现这个项目。`\n- Repository visit: `完整资源清单、每日检测状态和贡献入口都在 GitHub 仓库中维护。`\n- Browse resources: `普通用户也可以打开 zhuiju.me 直接查看今天可用的资源状态。`\n- Contribute: `发现好用的免费无广告资源，可以通过 GitHub Issue 推荐。`\n- Report failure: `如果某个资源失效，可以提交 Issue 帮助维护列表。`\n- Subscribe: `订阅频道获取资源更新通知。` Use only if the README confirms a channel.\n\n## Prohibited Claims\n\nDo not claim or imply:\n\n- The project hosts, stores, distributes, cracks, unlocks, or provides copyrighted video files.\n- All resources are legal, permanently valid, permanently safe, or available in every region.\n- Availability status is a universal guarantee. It only reflects the check environment and check time.\n- \"全网最全\", \"永久免费\", \"百分百无广告\", \"一定能看\", \"官方片源\", \"免会员破解\".\n- Any Star number, ranking, media recommendation, collection status, or user scale without current verification.\n- Paid promotion neutrality beyond what is documented. If saying \"不卖推广位\", verify it from current project text or user-provided instructions.\n\n## Compliance Boundary\n\nFor copyright-sensitive articles, include one concise note:\n\n> Awesome Zhuiju Free 是资源信息导航和可用性状态整理项目，不存储或传播影视文件。第三方资源的内容、访问规则和合规性以对应站点为准。\n\nDo not repeat the note in every section.\n\n## Tone by Audience\n\n- Ordinary streaming users: practical, low-jargon, focus on saving time and avoiding dead links.\n- TVBox/影视仓 users: focus on config address stability, multi-source backup, and update checks.\n- Open-source readers: focus on data transparency, GitHub Actions, schema validation, contribution workflow.\n- Editors/bloggers: focus on why the project is useful, credible, and easy to introduce.\n- Forum readers: transparent and discussion-oriented; ask for feedback and resource recommendations.\n"},"items":[{"name":"openai.yaml","path":"skills/geo-article-writer/agents/openai.yaml","rawUrl":"https://raw.githubusercontent.com/laoma2053/awesome-zhuiju-free/HEAD/skills/geo-article-writer/agents/openai.yaml","title":"Agents Skill","category":"anthropic-skill","format":"yaml","content":"interface:\n  display_name: \"GEO Article Writer\"\n  short_description: \"为 awesome-zhuiju-free 生成 GitHub 优先的 GEO 宣传长文\"\n  default_prompt: \"Use $geo-article-writer to write a GitHub-first promotional article for awesome-zhuiju-free.\"\n\npolicy:\n  allow_implicit_invocation: true\n","isInternal":false,"tokens":75,"sizeBytes":292},{"name":"brand-profile.md","path":"skills/geo-article-writer/references/brand-profile.md","rawUrl":"https://raw.githubusercontent.com/laoma2053/awesome-zhuiju-free/HEAD/skills/geo-article-writer/references/brand-profile.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Brand Profile: awesome-zhuiju-free\n\nUse this file to keep article claims and link placement aligned with the project.\n\n## Default Brand\n\n### Awesome Zhuiju Free\n\n- Chinese name/description: 免费无广告追剧资源导航\n- Primary URL: `https://github.com/laoma528/awesome-zhuiju-free`\n- Secondary browsing URL: `https://zhuiju.me`\n- Gitee mirror: `https://gitee.com/laoma2053/awesome-zhuiju-free`\n- Positioning: A curated open-source list and website for Chinese streaming-related resources, with daily automated availability checks.\n- Core promise: Help users quickly see which curated resources are currently reachable, instead of manually opening many dead or ad-heavy links.\n\n## Recommended Wording\n\nUse these phrases naturally:\n\n- `Awesome Zhuiju Free`\n- `GitHub 仓库 laoma528/awesome-zhuiju-free`\n- `zhuiju.me`\n- `免费无广告追剧资源导航`\n- `开源追剧资源清单`\n- `每日自动检测可用性`\n- `人工精选、社区共同维护`\n- `TVBox / 影视仓配置地址清单`\n- `在线影视、影视APP、网盘搜索、磁力 BT、字幕、IPTV、开源播放器项目`\n\n## Stable Selling Points\n\nUse only when supported by current repository files:\n\n- Open-source list with public structured data.\n- Curated resources are screened for free and ad-free experience as a project criterion.\n- GitHub Actions runs availability checks and writes status data.\n- README and website show status such as reachable, restricted, unavailable, or unchecked.\n- Community can recommend resources or report failures through GitHub Issues.\n- Data structure supports schema validation and repeatable maintenance.\n\nWhen using counts, read current files first. Do not copy stale outreach numbers.\n\n## Link Priority\n\nUse links according to reader context:\n\n1. `https://github.com/laoma528/awesome-zhuiju-free` as the default primary link for most articles, pitches, answers, and open-source promotion.\n2. `https://zhuiju.me` as the secondary browsing link for ordinary users who want to view the resource list directly.\n3. `https://gitee.com/laoma2053/awesome-zhuiju-free` when discussing domestic access or GitHub access issues.\n4. Telegram or other community links only after confirming the current README.\n\nIf a platform or article allows only one link, use the GitHub repository by default. Use `zhuiju.me` as the only link only for strongly non-technical contexts where direct browsing matters more than GitHub Star or contribution conversion.\n\n## CTA Patterns\n\nChoose one CTA, not several:\n\n- Star support: `如果这个清单对你有帮助，可以在 GitHub 点一个 Star，方便后续找到，也能帮助更多人发现这个项目。`\n- Repository visit: `完整资源清单、每日检测状态和贡献入口都在 GitHub 仓库中维护。`\n- Browse resources: `普通用户也可以打开 zhuiju.me 直接查看今天可用的资源状态。`\n- Contribute: `发现好用的免费无广告资源，可以通过 GitHub Issue 推荐。`\n- Report failure: `如果某个资源失效，可以提交 Issue 帮助维护列表。`\n- Subscribe: `订阅频道获取资源更新通知。` Use only if the README confirms a channel.\n\n## Prohibited Claims\n\nDo not claim or imply:\n\n- The project hosts, stores, distributes, cracks, unlocks, or provides copyrighted video files.\n- All resources are legal, permanently valid, permanently safe, or available in every region.\n- Availability status is a universal guarantee. It only reflects the check environment and check time.\n- \"全网最全\", \"永久免费\", \"百分百无广告\", \"一定能看\", \"官方片源\", \"免会员破解\".\n- Any Star number, ranking, media recommendation, collection status, or user scale without current verification.\n- Paid promotion neutrality beyond what is documented. If saying \"不卖推广位\", verify it from current project text or user-provided instructions.\n\n## Compliance Boundary\n\nFor copyright-sensitive articles, include one concise note:\n\n> Awesome Zhuiju Free 是资源信息导航和可用性状态整理项目，不存储或传播影视文件。第三方资源的内容、访问规则和合规性以对应站点为准。\n\nDo not repeat the note in every section.\n\n## Tone by Audience\n\n- Ordinary streaming users: practical, low-jargon, focus on saving time and avoiding dead links.\n- TVBox/影视仓 users: focus on config address stability, multi-source backup, and update checks.\n- Open-source readers: focus on data transparency, GitHub Actions, schema validation, contribution workflow.\n- Editors/bloggers: focus on why the project is useful, credible, and easy to introduce.\n- Forum readers: transparent and discussion-oriented; ask for feedback and resource recommendations.\n","isInternal":false,"tokens":1160,"sizeBytes":4687},{"name":"platform-rules.md","path":"skills/geo-article-writer/references/platform-rules.md","rawUrl":"https://raw.githubusercontent.com/laoma2053/awesome-zhuiju-free/HEAD/skills/geo-article-writer/references/platform-rules.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Platform Rules\n\nAdapt the same topic to the target platform. Do not merely change the title.\n\n## Self-hosted Blog / Typecho\n\nGoal: search traffic plus AI-search-friendly answers.\n\n- Use one H1 title, H2/H3 body.\n- Put the direct answer in the first 150 Chinese characters.\n- Include 3+ answer-first sections.\n- Add FAQ when it reflects real reader questions.\n- Include SEO Title, Meta Description, Slug, summary, tags, category, and image alt suggestions.\n- Suggest Article or BlogPosting JSON-LD; suggest FAQPage only when FAQ is substantial.\n- Include \"最后核验：YYYY-MM-DD\" for time-sensitive counts or platform facts.\n- Add 1-3 natural links, with the GitHub repository as the default primary link and zhuiju.me as a secondary browsing link.\n\n## Zhihu\n\nGoal: answer the question convincingly without looking like an ad.\n\n- Start with a direct judgment or practical answer.\n- Mention the project only after giving standalone value.\n- Usually use at most one external link in the answer body; use the GitHub repository by default unless the answer is purely for ordinary users who need direct browsing.\n- Avoid SEO metadata, Schema, and promotional slogans.\n- Use fewer headings than a blog article.\n- Include tradeoffs, limitations, and alternative methods.\n- Do not ask for Star unless the question itself is about open-source support.\n\n## Juejin / SegmentFault\n\nGoal: technical story, implementation logic, or growth review.\n\n- Prefer topics such as GitHub Actions availability checks, structured JSON data, schema validation, issue-driven workflow, README conversion, and open-source growth.\n- Explain the problem, design choices, implementation outline, and tradeoffs.\n- Include code snippets only when they are accurate and useful.\n- Mark example code as simplified if not copied from repository.\n- Include what did not work or what remains limited when writing a growth review.\n- End with a useful principle or next step, not just a project link.\n\n## CSDN / 51CTO / Blogyuan\n\nGoal: reproducible technical documentation.\n\n- State environment, files, workflow, and expected result.\n- Keep commands copyable and explain where to run them.\n- Include verification and rollback/repair advice when editing workflows.\n- For GitHub Actions or platform behavior, verify current docs if version-sensitive.\n- Use troubleshooting tables for failures.\n- Avoid filler background and non-technical promotion.\n\n## SSPai Matrix\n\nGoal: practical tool article for readers who value clean experience.\n\n- Start from a real everyday pain point.\n- Explain the workflow: open page, scan status, choose category, submit issue if needed.\n- Emphasize clean, open, transparent, and maintained.\n- Keep technical explanation short and readable.\n- Use screenshots: README first screen, resource table with status, website browsing view.\n- End with a concise boundary note, not aggressive CTA.\n\n## Appinn / 小众软件\n\nGoal: concise discovery post for clean tools.\n\n- Format like a software recommendation.\n- Lead with labels such as `免费`、`开源`、`Web`.\n- Focus on no-ad curation, daily checks, and direct use without registration.\n- Keep language restrained; do not ask for Star.\n- Include platform, price, website, GitHub, and tags if needed.\n\n## V2EX / NodeSeek / HostLoc / Forums\n\nGoal: discussion and feedback, not one-way advertising.\n\n- Use first-person project context only if true or supplied by the user.\n- Ask concrete questions: categories to add, README clarity, TVBox configs, maintainers.\n- Be transparent about copyright and availability limitations.\n- Avoid \"求 Star\", \"全网最全\", and repeated links.\n- Keep the tone direct and community-native.\n\n## WeChat Public Account\n\nGoal: readable article or editor-ready contribution.\n\n- Use short paragraphs and clear section headings.\n- Keep one CTA. Prefer GitHub Star, GitHub repository visit, or GitHub Issue contribution; use website browsing CTA only for non-technical resource/tool accounts.\n- Put references at the end if needed.\n- Avoid dense SEO keyword repetition.\n- For GitHub-oriented accounts, include the technical differentiation and growth angle.\n- For resource/tool accounts, focus on the user pain and daily availability check.\n\n## Xiaohongshu Long Note\n\nGoal: scannable note with save/share value.\n\n- Keep the main text around 600-1200 Chinese characters unless requested otherwise.\n- First 100 characters must state who it helps and what problem it solves.\n- Use practical labels: 免费追剧、无广告、TVBox配置、电视盒子、资源失效.\n- Provide cover text and 3-6 page carousel outline when requested.\n- Avoid dense external links; direct readers to search the project name or the GitHub repository name if link placement is limited. Mention zhuiju.me only as the direct browsing entry.\n\n## Blogger / Editor Pitch\n\nGoal: make forwarding or coverage easy.\n\n- Provide a short personal note plus a ready-to-publish blurb.\n- Include 3 facts: what it is, why it is different, where to see it.\n- Keep the ask lightweight.\n- Do not overstate Star growth, media coverage, or uniqueness.\n","isInternal":false,"tokens":1085,"sizeBytes":5074},{"name":"quality-checklist.md","path":"skills/geo-article-writer/references/quality-checklist.md","rawUrl":"https://raw.githubusercontent.com/laoma2053/awesome-zhuiju-free/HEAD/skills/geo-article-writer/references/quality-checklist.md","title":"References Skill","category":"anthropic-skill","format":"markdown","content":"# Publication Quality Checklist\n\nUse this before delivering final content. Fix issues directly.\n\n## Facts\n\n- [ ] Current resource count and category counts were read from repository files when mentioned.\n- [ ] Star counts, external collection status, Telegram handle, and platform submission status were verified when mentioned.\n- [ ] Dates use exact dates for time-sensitive facts.\n- [ ] External sources actually support the claims attached to them.\n- [ ] No invented tests, rankings, quotes, user numbers, or endorsements.\n- [ ] Availability-check claims explain that status reflects a check environment and check time.\n\n## Project Safety\n\n- [ ] The project is described as a navigation/list/status project, not a hosting or distribution service.\n- [ ] No wording implies piracy, cracking, bypassing payment, or guaranteed access.\n- [ ] \"Free/ad-free\" is expressed as a curation criterion or best-effort screening, not a permanent guarantee.\n- [ ] Copyright and third-party responsibility boundary is included where the topic is sensitive.\n- [ ] No third-party resource is promised to be legal, safe, or permanently available.\n\n## GEO Quality\n\n- [ ] The opening gives a direct answer or central conclusion.\n- [ ] There are at least 3 self-contained answer blocks.\n- [ ] Important definitions name the entity clearly and avoid vague pronouns.\n- [ ] The article covers likely follow-up questions.\n- [ ] FAQ items are real reader questions, not empty schema stuffing.\n- [ ] Unique value is present: workflow, comparison, table, checklist, example, or maintainable framework.\n\n## SEO Quality\n\n- [ ] Title matches the primary search intent.\n- [ ] H1/H2/H3 hierarchy is clean.\n- [ ] Main keyword appears naturally in title, opening, at least one heading, and conclusion when appropriate.\n- [ ] Meta Description can stand alone.\n- [ ] Slug is short, lowercase, and hyphenated for self-hosted posts.\n- [ ] Internal links use descriptive anchors.\n- [ ] External links are relevant and verified.\n\n## Naturalness\n\n- [ ] No empty opening like \"随着时代的发展\".\n- [ ] No mechanical \"首先、其次、再次、最后\" chain.\n- [ ] Section structures vary naturally.\n- [ ] No repeated transition phrases or repeated conclusions.\n- [ ] No exaggerated marketing words.\n- [ ] No fake personal experience.\n- [ ] Paragraph length supports reading on the target platform.\n\n## Platform Fit\n\n- [ ] Zhihu versions answer the question before mentioning the project.\n- [ ] Technical platform versions include implementation details and verification.\n- [ ] SSPai/Appinn versions prioritize clean user experience and practical use.\n- [ ] Forum versions invite discussion and avoid hard-selling.\n- [ ] WeChat versions use one clear CTA.\n- [ ] Xiaohongshu versions include cover/carousel guidance when needed.\n- [ ] SEO metadata is not mixed into platforms where it should not be pasted.\n\n## Final Delivery\n\n- [ ] The body can be copied directly into the target platform.\n- [ ] Metadata, title, tags, and summary match the article.\n- [ ] Markdown tables, links, and code blocks render correctly.\n- [ ] No internal analysis, prompt text, or model self-review remains.\n- [ ] No claim says the article is guaranteed to rank or be cited by AI search.\n","isInternal":false,"tokens":690,"sizeBytes":3228},{"name":"SKILL.md","path":"skills/geo-article-writer/SKILL.md","rawUrl":"https://raw.githubusercontent.com/laoma2053/awesome-zhuiju-free/HEAD/skills/geo-article-writer/SKILL.md","title":"geo-article-writer","category":"anthropic-skill","format":"markdown","content":"---\nname: geo-article-writer\ndescription: Write, rewrite, adapt, or review Chinese GEO/SEO long-form content that primarily promotes the GitHub repository laoma528/awesome-zhuiju-free, with zhuiju.me as a secondary browsing entry. Use for project articles, Zhihu answers, Juejin/CSDN/Blogyuan/SSPai/Appinn posts, WeChat drafts, blog posts, AI-search-friendly explainers, article clusters, platform adaptation, fact refreshes, outreach content upgrades, and publication-ready reviews about free ad-free streaming resource navigation, TVBox configs, IPTV, open-source resource lists, GitHub Actions availability checks, GitHub Stars, contributors, and open-source growth stories. Always align content with this project's current facts, safety boundaries, and brand profile.\n---\n\n# GEO Article Writer\n\nProduce publishable Chinese long-form content for `awesome-zhuiju-free`: topic research, article structure, full draft, GEO/SEO optimization, platform adaptation, and final review.\n\nThis skill is not a generic marketing writer. It should turn the project's real assets into useful articles that can be cited by search engines, AI search tools, community readers, and platform editors without overstating what the project provides.\n\n## Core Inputs\n\nExtract these from the user request:\n\n- Topic, keyword, draft, target question, or target platform\n- Desired mode: create, rewrite, review, platform-adapt, cluster, update\n- Target reader: ordinary streaming user, TVBox/影视仓 user, GitHub/open-source reader, technical operator, editor/blogger\n- Goal: search traffic, AI search visibility, community discussion, project introduction, contributor recruitment, or external submission\n- Required facts, links, screenshots, or existing outreach files\n- Forbidden claims, sensitive wording, length, tone, and publication date\n\nIf the user gives only a topic, default to `create`.\nIf the user gives a draft, default to `rewrite`.\nIf the user asks whether an article is good enough, default to `review` and deliver a revised final version.\nIf the user names a platform, adapt to that platform instead of producing a generic blog article.\n\nDefault output language is Simplified Chinese.\n\n## Project Context\n\nBefore writing project-related claims, prefer local project files over memory:\n\n- Read `README.md` for current positioning, resource count, category names, community links, badges, and disclaimer.\n- Read `resources/resources.json` or `resources/README.md` when category counts or resource types matter.\n- Read `outreach/*.md` when the task is about repurposing existing promotional copy.\n- Read `reports/availability.json` only when the article needs current availability examples or status data.\n- Read `CONTRIBUTING.md` when explaining community contribution.\n\nUse `references/brand-profile.md` for stable brand language and prohibited claims.\nUse `references/platform-rules.md` for platform-specific writing rules.\nUse `references/quality-checklist.md` before final delivery.\n\nDo not reuse outdated numbers from old outreach drafts. Recalculate or read current files when saying \"currently collected X resources\", category counts, Star counts, Telegram handle, or publication status.\n\n## Required Research\n\nUse web research whenever facts may have changed, including:\n\n- GitHub Stars, topics, trending, external collection status, or repository metadata\n- Platform rules, editorial requirements, submission URLs, or current community norms\n- Third-party tools, TVBox/影视仓 ecosystem details, IPTV project details, software versions, or policy-sensitive facts\n- Any claim using \"latest\", \"currently\", \"2026\", \"recently\", \"now\", or similar time-sensitive wording\n- User-provided URLs that need to be summarized or cited\n\nPure rewriting of supplied text can skip web research if no new facts are added.\n\nUse sources in this order:\n\n1. This repository and official project pages\n2. Official docs, project repos, platform guidelines, government/standard sources\n3. Reputable media, community posts, and high-quality technical articles\n4. User-provided materials\n\nDo not cite search result snippets, anonymous reposts, or low-quality aggregation pages as factual support.\n\n## Content Strategy\n\nEach article needs one primary intent:\n\n- **Definition**: what awesome-zhuiju-free is, who should use it\n- **Tutorial**: how to find resources, use TVBox configs, subscribe to updates, submit resources\n- **Comparison**: ordinary navigation sites vs this open-source live list\n- **Decision**: whether this route suits a specific reader\n- **List**: methods, tools, categories, resources, contribution ideas\n- **Technical**: GitHub Actions checks, JSON data, schema validation, issue-driven publishing\n- **Growth story**: how the project gained attention, what worked, what did not\n- **Community post**: ask for feedback, recruit maintainers, invite resource recommendations\n\nSet a clear center judgment before writing. The article should answer what the reader can do next, not only introduce the project.\n\n## Writing Rules\n\n### Opening\n\nWithin the first 150 Chinese characters:\n\n- Answer the core question or state the central conclusion\n- Identify who the article is for\n- Preview the practical value\n\nAvoid empty openers such as \"随着互联网的发展\", \"在当今数字化时代\", \"众所周知\", \"本文将深入探讨\", and long suspense setups.\n\n### GEO Answer Blocks\n\nInclude at least 3 self-contained answer blocks in normal article sections. Each block should:\n\n- Directly answer a likely user question\n- Name the concrete entity, such as `Awesome Zhuiju Free`, `GitHub 仓库 laoma528/awesome-zhuiju-free`, `TVBox 配置地址`\n- Include conditions, limitations, or time scope when relevant\n- Be understandable without reading the surrounding article\n- Avoid vague pronouns like \"这个\", \"它\", \"上述\"\n\nDo not label them as \"AI 引用块\".\n\n### Style\n\n- Write answer-first, then explain.\n- Use concrete nouns and actions.\n- Keep paragraphs varied and readable.\n- Use tables only when comparison or scanning is improved.\n- Avoid mechanical \"首先、其次、最后\" chains.\n- Avoid exaggerated marketing words such as \"全网最全\", \"永久免费可用\", \"神器\", \"闭眼入\", \"百分百\".\n- Do not invent personal experience, user testimonials, editor endorsements, rankings, tests, or traffic numbers.\n\n### Project Claim Boundaries\n\nAlways observe these boundaries:\n\n- The project is a curated open-source navigation/list, not a content host.\n- Do not say the project stores, distributes, cracks, or provides copyrighted video files.\n- Do not promise that third-party resources are always available, legal, safe, ad-free forever, or available in every region.\n- Say availability checks reflect the GitHub Actions environment at the check time.\n- Explain that \"free/ad-free\" is a curation criterion and best-effort manual screening, not a legal guarantee or permanent guarantee.\n- For copyright-sensitive contexts, include one concise boundary note near the relevant section or at the end.\n\n## Brand and Link Use\n\nRead `references/brand-profile.md` before adding project links.\n\nDefault link priority:\n\n1. GitHub: `https://github.com/laoma528/awesome-zhuiju-free`\n2. Website: `https://zhuiju.me`\n3. Gitee mirror: `https://gitee.com/laoma2053/awesome-zhuiju-free`\n4. Telegram channel only if current README confirms it\n\nUse links only where they help the reader. A normal article should mention the project 1-3 times, unless the whole article is a project announcement.\nWhen choosing only one primary CTA, prefer the GitHub repository unless the target platform is aimed at non-technical users who mainly need a direct browsing page.\n\nPrefer descriptive anchors:\n\n- `免费无广告追剧资源导航`\n- `查看每日检测状态`\n- `提交新资源或报告失效`\n- `TVBox/影视仓配置地址清单`\n\nAvoid naked repeated links, \"点这里\", and anchors unrelated to the target page.\n\n## Platform Adaptation\n\nRead `references/platform-rules.md` and adapt structure, not just title.\n\nCommon defaults:\n\n- **Self-hosted blog / Typecho**: complete SEO metadata, FAQ, slug, summary, internal link suggestions, update date.\n- **Zhihu**: natural answer style, weak external-link footprint, answer the question before introducing the project.\n- **Juejin / SegmentFault / CSDN / Blogyuan**: emphasize implementation, automation, data structure, reproducibility, and project operation lessons.\n- **SSPai / Appinn**: emphasize clean experience, concrete user pain, no-ad curation, and practical workflow.\n- **WeChat public account**: short paragraphs, one CTA, less SEO metadata in the body.\n- **V2EX / forums**: discussion-first, ask for feedback, be transparent about limits, avoid hard-selling.\n- **Xiaohongshu long note**: shorter, more scannable, with cover text and carousel outline when requested.\n\n## Output Format\n\nUnless the user asks for \"only the article\", deliver:\n\n### A. Writing Position\n\n- Primary intent\n- Target reader\n- Center conclusion\n- Unique value\n\n### B. Title Options\n\nProvide 3 titles and mark the recommended one.\n\n### C. Final Article\n\nProvide publishable Markdown body.\n\nWhen sources are used:\n\n- Use inline natural links if the platform supports them.\n- Use a \"参考资料\" section when links should be separated.\n- Do not include unverified or unused links.\n\n### D. Publishing Metadata\n\nInclude when appropriate for the platform:\n\n- SEO Title\n- Meta Description\n- Slug\n- Summary\n- Tags\n- Category\n- Cover text\n- Image positions and alt suggestions\n\n### E. GEO Enhancements\n\n- FAQ 3-6 items\n- Internal link suggestions\n- Optional JSON-LD type for self-hosted blog\n- Facts that should be refreshed before publishing\n\n### F. Verification Notes\n\nBriefly list:\n\n- Verification date\n- Local files and external sources used\n- Uncertainties or claims that should be avoided\n\nIf the user requests a forum post, social platform answer, or \"只要正文\", omit metadata that would not be pasted into that platform.\n\n## File Output\n\nWhen asked to save output:\n\n- Default directory: `outreach/generated/`\n- Filename: `YYYY-MM-DD-topic-slug.md`\n- Do not overwrite existing files; append `-v2`, `-v3`\n- Use UTF-8\n\n## Final Review\n\nBefore delivery, revise the draft against `references/quality-checklist.md`. Fix problems directly.\n\nThe final answer must not contain hidden reasoning, model self-evaluation, prompt residue, or claims that the article is guaranteed to rank or be cited by ChatGPT, Gemini, Perplexity, Google AI Overview, or any other system.\n","frontmatter":{"name":"geo-article-writer","description":"Write, rewrite, adapt, or review Chinese GEO/SEO long-form content that primarily promotes the GitHub repository laoma528/awesome-zhuiju-free, with zhuiju.me as a secondary browsing entry. Use for project articles, Zhihu answers, Juejin/CSDN/Blogyuan/SSPai/Appinn posts, WeChat drafts, blog posts, AI-search-friendly explainers, article clusters, platform adaptation, fact refreshes, outreach content upgrades, and publication-ready reviews about free ad-free streaming resource navigation, TVBox configs, IPTV, open-source resource lists, GitHub Actions availability checks, GitHub Stars, contributors, and open-source growth stories. Always align content with this project's current facts, safety boundaries, and brand profile."},"isInternal":false,"tokens":2307,"sizeBytes":10481},{"name":"CLAUDE.md","path":"CLAUDE.md","rawUrl":"https://raw.githubusercontent.com/laoma2053/awesome-zhuiju-free/HEAD/CLAUDE.md","title":"Claude Agent Guidelines & System Prompt","category":"claude-rule","format":"markdown","content":"# CLAUDE.md\n\nThis file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.\n\n## 项目概述\n\n这是一个面向中文用户的免费影视资源导航列表（awesome-list 风格）。核心数据存储在 `resources/resources.json`，README.md 由脚本从数据自动生成，不应手动编辑 README 中的精选资源区块。\n\n## 常用命令\n\n所有脚本均为原生 Node.js ESM 模块，**无需 `npm install`**，直接运行：\n\n```bash\n# 验证所有 JSON 数据格式合法性\nnode scripts/validate-data.mjs\n\n# 将 resources.json 中的精选资源同步到 README.md（重建表格）\nnode scripts/sync-readme.mjs\n\n# 检测所有精选资源的 URL 可访问性，更新 reports/availability.json 和 README\nnode scripts/check-availability.mjs\n\n# 更新 open_source 类型资源的 GitHub star 数和最近推送时间\nGITHUB_TOKEN=<token> node scripts/update-open-source-stats.mjs\n```\n\nCI 使用 Node.js 24；本地需 Node.js 18+（使用了原生 `fetch` 和顶层 `await`）。\n\n## 架构与数据流\n\n### 单一数据源\n\n`resources/resources.json` 是所有资源的唯一数据源。README 中 `<!-- featured-resources:start -->` 和 `<!-- featured-resources:end -->` 之间的全部内容，以及 `<!-- resource-count:start/end -->` 徽章，**均由 `scripts/sync-readme.mjs` 自动覆盖**，手动编辑这些区块会在下次 CI 运行时被还原。\n\n### 脚本职责\n\n| 脚本 | 输入 | 输出 |\n|---|---|---|\n| `sync-readme.mjs` | `resources/resources.json`，`reports/availability.json` | `README.md`（精选区块） |\n| `check-availability.mjs` | `resources/resources.json`，`README.md` | `reports/availability.json`，`README.md`（状态标记） |\n| `validate-data.mjs` | 全部 JSON 文件 | 仅 stdout 错误报告 |\n| `update-open-source-stats.mjs` | `resources/resources.json`（GitHub API） | `resources/resources.json` |\n\n### 自动化流程\n\n- **每日 09:00 (北京时间)**：`check-availability.yml` 依次执行 sync-readme → check-availability → validate-data，提交结果\n- **PR / push to main**：`validate-data.yml` 校验数据格式\n- **Issue 审核**：维护者在资源推荐 Issue 下评论 `ok` 触发 `issue-to-resource-pr.yml`，自动将资源写入 resources.json 并发布\n\n## resources.json 数据约束\n\n新增或修改资源时必须满足以下规则（`validate-data.mjs` 会检测）：\n\n- `id`：小写字母和数字，以 `-` 分隔，模式 `^[a-z0-9]+(?:-[a-z0-9]+)*$`，全局唯一\n- `url`：必须是 HTTPS，`tvbox_config` 类型例外可用 HTTP\n- `category`：必须是枚举值之一：`online_video` / `video_app` / `cloud_search` / `magnet_search` / `subtitles` / `player` / `subscription` / `tvbox_config` / `membership` / `open_source` / `other`\n- `scores.*`：`more` / `speed` / `clean` / `stable` / `ease` 均为 1.0–5.0，最多一位小数\n- `risks.*`：`copyright` / `safety` / `privacy` / `payment` 枚举值：`low` / `medium` / `high` / `unknown`\n- `verification.status`：`pending` / `verified` / `recommended` / `caution` / `temporarily_unavailable` / `removed`\n- `featured: true` 的资源必须提供 `summary_short`（40 个字符以内）\n- `category: \"open_source\"` 的资源必须包含 `github.full_name`（`owner/repo` 格式）、`github.stars`（非负整数）、`github.pushed_at`（ISO 时间戳）\n- 日期字段格式：`YYYY-MM-DD`\n\n修改 resources.json 后，手动运行 `node scripts/sync-readme.mjs` 可在本地预览 README 变化。\n","isInternal":false,"tokens":1096,"sizeBytes":3506}],"systemPromptSnippet":"<yakaai_skills repo=\"laoma2053/awesome-zhuiju-free\">\n<!-- File: skills/geo-article-writer/agents/openai.yaml (Tokens: ~75 | Category: anthropic-skill) -->\ninterface:\n  display_name: \"GEO Article Writer\"\n  short_description: \"为 awesome-zhuiju-free 生成 GitHub 优先的 GEO 宣传长文\"\n  default_prompt: \"Use $geo-article-writer to write a GitHub-first promotional article for awesome-zhuiju-free.\"\n\npolicy:\n  allow_implicit_invocation: true\n\n<!-- File: skills/geo-article-writer/references/brand-profile.md (Tokens: ~1160 | Category: anthropic-skill) -->\n# Brand Profile: awesome-zhuiju-free\n\nUse this file to keep article claims and link placement aligned with the project.\n\n## Default Brand\n\n### Awesome Zhuiju Free\n\n- Chinese name/description: 免费无广告追剧资源导航\n- Primary URL: `https://github.com/laoma528/awesome-zhuiju-free`\n- Secondary browsing URL: `https://zhuiju.me`\n- Gitee mirror: `https://gitee.com/laoma2053/awesome-zhuiju-free`\n- Positioning: A curated open-source list and website for Chinese streaming-related resources, with daily automated availability checks.\n- Core promise: Help users quickly see which curated resources are currently reachable, instead of manually opening many dead or ad-heavy links.\n\n## Recommended Wording\n\nUse these phrases naturally:\n\n- `Awesome Zhuiju Free`\n- `GitHub 仓库 laoma528/awesome-zhuiju-free`\n- `zhuiju.me`\n- `免费无广告追剧资源导航`\n- `开源追剧资源清单`\n- `每日自动检测可用性`\n- `人工精选、社区共同维护`\n- `TVBox / 影视仓配置地址清单`\n- `在线影视、影视APP、网盘搜索、磁力 BT、字幕、IPTV、开源播放器项目`\n\n## Stable Selling Points\n\nUse only when supported by current repository files:\n\n- Open-source list with public structured data.\n- Curated resources are screened for free and ad-free experience as a project criterion.\n- GitHub Actions runs availability checks and writes status data.\n- README and website show status such as reachable, restricted, unavailable, or unchecked.\n- Community can recommend resources or report failures through GitHub Issues.\n- Data structure supports schema validation and repeatable maintenance.\n\nWhen using counts, read current files first. Do not copy stale outreach numbers.\n\n## Link Priority\n\nUse links according to reader context:\n\n1. `https://github.com/laoma528/awesome-zhuiju-free` as the default primary link for most articles, pitches, answers, and open-source promotion.\n2. `https://zhuiju.me` as the secondary browsing link for ordinary users who want to view the resource list directly.\n3. `https://gitee.com/laoma2053/awesome-zhuiju-free` when discussing domestic access or GitHub access issues.\n4. Telegram or other community links only after confirming the current README.\n\nIf a platform or article allows only one link, use the GitHub repository by default. Use `zhuiju.me` as the only link only for strongly non-technical contexts where direct browsing matters more than GitHub Star or contribution conversion.\n\n## CTA Patterns\n\nChoose one CTA, not several:\n\n- Star support: `如果这个清单对你有帮助，可以在 GitHub 点一个 Star，方便后续找到，也能帮助更多人发现这个项目。`\n- Repository visit: `完整资源清单、每日检测状态和贡献入口都在 GitHub 仓库中维护。`\n- Browse resources: `普通用户也可以打开 zhuiju.me 直接查看今天可用的资源状态。`\n- Contribute: `发现好用的免费无广告资源，可以通过 GitHub Issue 推荐。`\n- Report failure: `如果某个资源失效，可以提交 Issue 帮助维护列表。`\n- Subscribe: `订阅频道获取资源更新通知。` Use only if the README confirms a channel.\n\n## Prohibited Claims\n\nDo not claim or imply:\n\n- The project hosts, stores, distributes, cracks, unlocks, or provides copyrighted video files.\n- All resources are legal, permanently valid, permanently safe, or available in every region.\n- Availability status is a universal guarantee. It only reflects the check environment and check time.\n- \"全网最全\", \"永久免费\", \"百分百无广告\", \"一定能看\", \"官方片源\", \"免会员破解\".\n- Any Star number, ranking, media recommendation, collection status, or user scale without current verification.\n- Paid promotion neutrality beyond what is documented. If saying \"不卖推广位\", verify it from current project text or user-provided instructions.\n\n## Compliance Boundary\n\nFor copyright-sensitive articles, include one concise note:\n\n> Awesome Zhuiju Free 是资源信息导航和可用性状态整理项目，不存储或传播影视文件。第三方资源的内容、访问规则和合规性以对应站点为准。\n\nDo not repeat the note in every section.\n\n## Tone by Audience\n\n- Ordinary streaming users: practical, low-jargon, focus on saving time and avoiding dead links.\n- TVBox/影视仓 users: focus on config address stability, multi-source backup, and update checks.\n- Open-source readers: focus on data transparency, GitHub Actions, schema validation, contribution workflow.\n- Editors/bloggers: focus on why the project is useful, credible, and easy to introduce.\n- Forum readers: transparent and discussion-oriented; ask for feedback and resource recommendations.\n\n<!-- File: skills/geo-article-writer/references/platform-rules.md (Tokens: ~1085 | Category: anthropic-skill) -->\n# Platform Rules\n\nAdapt the same topic to the target platform. Do not merely change the title.\n\n## Self-hosted Blog / Typecho\n\nGoal: search traffic plus AI-search-friendly answers.\n\n- Use one H1 title, H2/H3 body.\n- Put the direct answer in the first 150 Chinese characters.\n- Include 3+ answer-first sections.\n- Add FAQ when it reflects real reader questions.\n- Include SEO Title, Meta Description, Slug, summary, tags, category, and image alt suggestions.\n- Suggest Article or BlogPosting JSON-LD; suggest FAQPage only when FAQ is substantial.\n- Include \"最后核验：YYYY-MM-DD\" for time-sensitive counts or platform facts.\n- Add 1-3 natural links, with the GitHub repository as the default primary link and zhuiju.me as a secondary browsing link.\n\n## Zhihu\n\nGoal: answer the question convincingly without looking like an ad.\n\n- Start with a direct judgment or practical answer.\n- Mention the project only after giving standalone value.\n- Usually use at most one external link in the answer body; use the GitHub repository by default unless the answer is purely for ordinary users who need direct browsing.\n- Avoid SEO metadata, Schema, and promotional slogans.\n- Use fewer headings than a blog article.\n- Include tradeoffs, limitations, and alternative methods.\n- Do not ask for Star unless the question itself is about open-source support.\n\n## Juejin / SegmentFault\n\nGoal: technical story, implementation logic, or growth review.\n\n- Prefer topics such as GitHub Actions availability checks, structured JSON data, schema validation, issue-driven workflow, README conversion, and open-source growth.\n- Explain the problem, design choices, implementation outline, and tradeoffs.\n- Include code snippets only when they are accurate and useful.\n- Mark example code as simplified if not copied from repository.\n- Include what did not work or what remains limited when writing a growth review.\n- End with a useful principle or next step, not just a project link.\n\n## CSDN / 51CTO / Blogyuan\n\nGoal: reproducible technical documentation.\n\n- State environment, files, workflow, and expected result.\n- Keep commands copyable and explain where to run them.\n- Include verification and rollback/repair advice when editing workflows.\n- For GitHub Actions or platform behavior, verify current docs if version-sensitive.\n- Use troubleshooting tables for failures.\n- Avoid filler background and non-technical promotion.\n\n## SSPai Matrix\n\nGoal: practical tool article for readers who value clean experience.\n\n- Start from a real everyday pain point.\n- Explain the workflow: open page, scan status, choose category, submit issue if needed.\n- Emphasize clean, open, transparent, and maintained.\n- Keep technical explanation short and readable.\n- Use screenshots: README first screen, resource table with status, website browsing view.\n- End with a concise boundary note, not aggressive CTA.\n\n## Appinn / 小众软件\n\nGoal: concise discovery post for clean tools.\n\n- Format like a software recommendation.\n- Lead with labels such as `免费`、`开源`、`Web`.\n- Focus on no-ad curation, daily checks, and direct use without registration.\n- Keep language restrained; do not ask for Star.\n- Include platform, price, website, GitHub, and tags if needed.\n\n## V2EX / NodeSeek / HostLoc / Forums\n\nGoal: discussion and feedback, not one-way advertising.\n\n- Use first-person project context only if true or supplied by the user.\n- Ask concrete questions: categories to add, README clarity, TVBox configs, maintainers.\n- Be transparent about copyright and availability limitations.\n- Avoid \"求 Star\", \"全网最全\", and repeated links.\n- Keep the tone direct and community-native.\n\n## WeChat Public Account\n\nGoal: readable article or editor-ready contribution.\n\n- Use short paragraphs and clear section headings.\n- Keep one CTA. Prefer GitHub Star, GitHub repository visit, or GitHub Issue contribution; use website browsing CTA only for non-technical resource/tool accounts.\n- Put references at the end if needed.\n- Avoid dense SEO keyword repetition.\n- For GitHub-oriented accounts, include the technical differentiation and growth angle.\n- For resource/tool accounts, focus on the user pain and daily availability check.\n\n## Xiaohongshu Long Note\n\nGoal: scannable note with save/share value.\n\n- Keep the main text around 600-1200 Chinese characters unless requested otherwise.\n- First 100 characters must state who it helps and what problem it solves.\n- Use practical labels: 免费追剧、无广告、TVBox配置、电视盒子、资源失效.\n- Provide cover text and 3-6 page carousel outline when requested.\n- Avoid dense external links; direct readers to search the project name or the GitHub repository name if link placement is limited. Mention zhuiju.me only as the direct browsing entry.\n\n## Blogger / Editor Pitch\n\nGoal: make forwarding or coverage easy.\n\n- Provide a short personal note plus a ready-to-publish blurb.\n- Include 3 facts: what it is, why it is different, where to see it.\n- Keep the ask lightweight.\n- Do not overstate Star growth, media coverage, or uniqueness.\n\n<!-- File: skills/geo-article-writer/references/quality-checklist.md (Tokens: ~690 | Category: anthropic-skill) -->\n# Publication Quality Checklist\n\nUse this before delivering final content. Fix issues directly.\n\n## Facts\n\n- [ ] Current resource count and category counts were read from repository files when mentioned.\n- [ ] Star counts, external collection status, Telegram handle, and platform submission status were verified when mentioned.\n- [ ] Dates use exact dates for time-sensitive facts.\n- [ ] External sources actually support the claims attached to them.\n- [ ] No invented tests, rankings, quotes, user numbers, or endorsements.\n- [ ] Availability-check claims explain that status reflects a check environment and check time.\n\n## Project Safety\n\n- [ ] The project is described as a navigation/list/status project, not a hosting or distribution service.\n- [ ] No wording implies piracy, cracking, bypassing payment, or guaranteed access.\n- [ ] \"Free/ad-free\" is expressed as a curation criterion or best-effort screening, not a permanent guarantee.\n- [ ] Copyright and third-party responsibility boundary is included where the topic is sensitive.\n- [ ] No third-party resource is promised to be legal, safe, or permanently available.\n\n## GEO Quality\n\n- [ ] The opening gives a direct answer or central conclusion.\n- [ ] There are at least 3 self-contained answer blocks.\n- [ ] Important definitions name the entity clearly and avoid vague pronouns.\n- [ ] The article covers likely follow-up questions.\n- [ ] FAQ items are real reader questions, not empty schema stuffing.\n- [ ] Unique value is present: workflow, comparison, table, checklist, example, or maintainable framework.\n\n## SEO Quality\n\n- [ ] Title matches the primary search intent.\n- [ ] H1/H2/H3 hierarchy is clean.\n- [ ] Main keyword appears naturally in title, opening, at least one heading, and conclusion when appropriate.\n- [ ] Meta Description can stand alone.\n- [ ] Slug is short, lowercase, and hyphenated for self-hosted posts.\n- [ ] Internal links use descriptive anchors.\n- [ ] External links are relevant and verified.\n\n## Naturalness\n\n- [ ] No empty opening like \"随着时代的发展\".\n- [ ] No mechanical \"首先、其次、再次、最后\" chain.\n- [ ] Section structures vary naturally.\n- [ ] No repeated transition phrases or repeated conclusions.\n- [ ] No exaggerated marketing words.\n- [ ] No fake personal experience.\n- [ ] Paragraph length supports reading on the target platform.\n\n## Platform Fit\n\n- [ ] Zhihu versions answer the question before mentioning the project.\n- [ ] Technical platform versions include implementation details and verification.\n- [ ] SSPai/Appinn versions prioritize clean user experience and practical use.\n- [ ] Forum versions invite discussion and avoid hard-selling.\n- [ ] WeChat versions use one clear CTA.\n- [ ] Xiaohongshu versions include cover/carousel guidance when needed.\n- [ ] SEO metadata is not mixed into platforms where it should not be pasted.\n\n## Final Delivery\n\n- [ ] The body can be copied directly into the target platform.\n- [ ] Metadata, title, tags, and summary match the article.\n- [ ] Markdown tables, links, and code blocks render correctly.\n- [ ] No internal analysis, prompt text, or model self-review remains.\n- [ ] No claim says the article is guaranteed to rank or be cited by AI search.\n\n<!-- File: skills/geo-article-writer/SKILL.md (Tokens: ~2307 | Category: anthropic-skill) -->\n---\nname: geo-article-writer\ndescription: Write, rewrite, adapt, or review Chinese GEO/SEO long-form content that primarily promotes the GitHub repository laoma528/awesome-zhuiju-free, with zhuiju.me as a secondary browsing entry. Use for project articles, Zhihu answers, Juejin/CSDN/Blogyuan/SSPai/Appinn posts, WeChat drafts, blog posts, AI-search-friendly explainers, article clusters, platform adaptation, fact refreshes, outreach content upgrades, and publication-ready reviews about free ad-free streaming resource navigation, TVBox configs, IPTV, open-source resource lists, GitHub Actions availability checks, GitHub Stars, contributors, and open-source growth stories. Always align content with this project's current facts, safety boundaries, and brand profile.\n---\n\n# GEO Article Writer\n\nProduce publishable Chinese long-form content for `awesome-zhuiju-free`: topic research, article structure, full draft, GEO/SEO optimization, platform adaptation, and final review.\n\nThis skill is not a generic marketing writer. It should turn the project's real assets into useful articles that can be cited by search engines, AI search tools, community readers, and platform editors without overstating what the project provides.\n\n## Core Inputs\n\nExtract these from the user request:\n\n- Topic, keyword, draft, target question, or target platform\n- Desired mode: create, rewrite, review, platform-adapt, cluster, update\n- Target reader: ordinary streaming user, TVBox/影视仓 user, GitHub/open-source reader, technical operator, editor/blogger\n- Goal: search traffic, AI search visibility, community discussion, project introduction, contributor recruitment, or external submission\n- Required facts, links, screenshots, or existing outreach files\n- Forbidden claims, sensitive wording, length, tone, and publication date\n\nIf the user gives only a topic, default to `create`.\nIf the user gives a draft, default to `rewrite`.\nIf the user asks whether an article is good enough, default to `review` and deliver a revised final version.\nIf the user names a platform, adapt to that platform instead of producing a generic blog article.\n\nDefault output language is Simplified Chinese.\n\n## Project Context\n\nBefore writing project-related claims, prefer local project files over memory:\n\n- Read `README.md` for current positioning, resource count, category names, community links, badges, and disclaimer.\n- Read `resources/resources.json` or `resources/README.md` when category counts or resource types matter.\n- Read `outreach/*.md` when the task is about repurposing existing promotional copy.\n- Read `reports/availability.json` only when the article needs current availability examples or status data.\n- Read `CONTRIBUTING.md` when explaining community contribution.\n\nUse `references/brand-profile.md` for stable brand language and prohibited claims.\nUse `references/platform-rules.md` for platform-specific writing rules.\nUse `references/quality-checklist.md` before final delivery.\n\nDo not reuse outdated numbers from old outreach drafts. Recalculate or read current files when saying \"currently collected X resources\", category counts, Star counts, Telegram handle, or publication status.\n\n## Required Research\n\nUse web research whenever facts may have changed, including:\n\n- GitHub Stars, topics, trending, external collection status, or repository metadata\n- Platform rules, editorial requirements, submission URLs, or current community norms\n- Third-party tools, TVBox/影视仓 ecosystem details, IPTV project details, software versions, or policy-sensitive facts\n- Any claim using \"latest\", \"currently\", \"2026\", \"recently\", \"now\", or similar time-sensitive wording\n- User-provided URLs that need to be summarized or cited\n\nPure rewriting of supplied text can skip web research if no new facts are added.\n\nUse sources in this order:\n\n1. This repository and official project pages\n2. Official docs, project repos, platform guidelines, government/standard sources\n3. Reputable media, community posts, and high-quality technical articles\n4. User-provided materials\n\nDo not cite search result snippets, anonymous reposts, or low-quality aggregation pages as factual support.\n\n## Content Strategy\n\nEach article needs one primary intent:\n\n- **Definition**: what awesome-zhuiju-free is, who should use it\n- **Tutorial**: how to find resources, use TVBox configs, subscribe to updates, submit resources\n- **Comparison**: ordinary navigation sites vs this open-source live list\n- **Decision**: whether this route suits a specific reader\n- **List**: methods, tools, categories, resources, contribution ideas\n- **Technical**: GitHub Actions checks, JSON data, schema validation, issue-driven publishing\n- **Growth story**: how the project gained attention, what worked, what did not\n- **Community post**: ask for feedback, recruit maintainers, invite resource recommendations\n\nSet a clear center judgment before writing. The article should answer what the reader can do next, not only introduce the project.\n\n## Writing Rules\n\n### Opening\n\nWithin the first 150 Chinese characters:\n\n- Answer the core question or state the central conclusion\n- Identify who the article is for\n- Preview the practical value\n\nAvoid empty openers such as \"随着互联网的发展\", \"在当今数字化时代\", \"众所周知\", \"本文将深入探讨\", and long suspense setups.\n\n### GEO Answer Blocks\n\nInclude at least 3 self-contained answer blocks in normal article sections. Each block should:\n\n- Directly answer a likely user question\n- Name the concrete entity, such as `Awesome Zhuiju Free`, `GitHub 仓库 laoma528/awesome-zhuiju-free`, `TVBox 配置地址`\n- Include conditions, limitations, or time scope when relevant\n- Be understandable without reading the surrounding article\n- Avoid vague pronouns like \"这个\", \"它\", \"上述\"\n\nDo not label them as \"AI 引用块\".\n\n### Style\n\n- Write answer-first, then explain.\n- Use concrete nouns and actions.\n- Keep paragraphs varied and readable.\n- Use tables only when comparison or scanning is improved.\n- Avoid mechanical \"首先、其次、最后\" chains.\n- Avoid exaggerated marketing words such as \"全网最全\", \"永久免费可用\", \"神器\", \"闭眼入\", \"百分百\".\n- Do not invent personal experience, user testimonials, editor endorsements, rankings, tests, or traffic numbers.\n\n### Project Claim Boundaries\n\nAlways observe these boundaries:\n\n- The project is a curated open-source navigation/list, not a content host.\n- Do not say the project stores, distributes, cracks, or provides copyrighted video files.\n- Do not promise that third-party resources are always available, legal, safe, ad-free forever, or available in every region.\n- Say availability checks reflect the GitHub Actions environment at the check time.\n- Explain that \"free/ad-free\" is a curation criterion and best-effort manual screening, not a legal guarantee or permanent guarantee.\n- For copyright-sensitive contexts, include one concise boundary note near the relevant section or at the end.\n\n## Brand and Link Use\n\nRead `references/brand-profile.md` before adding project links.\n\nDefault link priority:\n\n1. GitHub: `https://github.com/laoma528/awesome-zhuiju-free`\n2. Website: `https://zhuiju.me`\n3. Gitee mirror: `https://gitee.com/laoma2053/awesome-zhuiju-free`\n4. Telegram channel only if current README confirms it\n\nUse links only where they help the reader. A normal article should mention the project 1-3 times, unless the whole article is a project announcement.\nWhen choosing only one primary CTA, prefer the GitHub repository unless the target platform is aimed at non-technical users who mainly need a direct browsing page.\n\nPrefer descriptive anchors:\n\n- `免费无广告追剧资源导航`\n- `查看每日检测状态`\n- `提交新资源或报告失效`\n- `TVBox/影视仓配置地址清单`\n\nAvoid naked repeated links, \"点这里\", and anchors unrelated to the target page.\n\n## Platform Adaptation\n\nRead `references/platform-rules.md` and adapt structure, not just title.\n\nCommon defaults:\n\n- **Self-hosted blog / Typecho**: complete SEO metadata, FAQ, slug, summary, internal link suggestions, update date.\n- **Zhihu**: natural answer style, weak external-link footprint, answer the question before introducing the project.\n- **Juejin / SegmentFault / CSDN / Blogyuan**: emphasize implementation, automation, data structure, reproducibility, and project operation lessons.\n- **SSPai / Appinn**: emphasize clean experience, concrete user pain, no-ad curation, and practical workflow.\n- **WeChat public account**: short paragraphs, one CTA, less SEO metadata in the body.\n- **V2EX / forums**: discussion-first, ask for feedback, be transparent about limits, avoid hard-selling.\n- **Xiaohongshu long note**: shorter, more scannable, with cover text and carousel outline when requested.\n\n## Output Format\n\nUnless the user asks for \"only the article\", deliver:\n\n### A. Writing Position\n\n- Primary intent\n- Target reader\n- Center conclusion\n- Unique value\n\n### B. Title Options\n\nProvide 3 titles and mark the recommended one.\n\n### C. Final Article\n\nProvide publishable Markdown body.\n\nWhen sources are used:\n\n- Use inline natural links if the platform supports them.\n- Use a \"参考资料\" section when links should be separated.\n- Do not include unverified or unused links.\n\n### D. Publishing Metadata\n\nInclude when appropriate for the platform:\n\n- SEO Title\n- Meta Description\n- Slug\n- Summary\n- Tags\n- Category\n- Cover text\n- Image positions and alt suggestions\n\n### E. GEO Enhancements\n\n- FAQ 3-6 items\n- Internal link suggestions\n- Optional JSON-LD type for self-hosted blog\n- Facts that should be refreshed before publishing\n\n### F. Verification Notes\n\nBriefly list:\n\n- Verification date\n- Local files and external sources used\n- Uncertainties or claims that should be avoided\n\nIf the user requests a forum post, social platform answer, or \"只要正文\", omit metadata that would not be pasted into that platform.\n\n## File Output\n\nWhen asked to save output:\n\n- Default directory: `outreach/generated/`\n- Filename: `YYYY-MM-DD-topic-slug.md`\n- Do not overwrite existing files; append `-v2`, `-v3`\n- Use UTF-8\n\n## Final Review\n\nBefore delivery, revise the draft against `references/quality-checklist.md`. Fix problems directly.\n\nThe final answer must not contain hidden reasoning, model self-evaluation, prompt residue, or claims that the article is guaranteed to rank or be cited by ChatGPT, Gemini, Perplexity, Google AI Overview, or any other system.\n\n<!-- File: CLAUDE.md (Tokens: ~1096 | Category: claude-rule) -->\n# CLAUDE.md\n\nThis file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.\n\n## 项目概述\n\n这是一个面向中文用户的免费影视资源导航列表（awesome-list 风格）。核心数据存储在 `resources/resources.json`，README.md 由脚本从数据自动生成，不应手动编辑 README 中的精选资源区块。\n\n## 常用命令\n\n所有脚本均为原生 Node.js ESM 模块，**无需 `npm install`**，直接运行：\n\n```bash\n# 验证所有 JSON 数据格式合法性\nnode scripts/validate-data.mjs\n\n# 将 resources.json 中的精选资源同步到 README.md（重建表格）\nnode scripts/sync-readme.mjs\n\n# 检测所有精选资源的 URL 可访问性，更新 reports/availability.json 和 README\nnode scripts/check-availability.mjs\n\n# 更新 open_source 类型资源的 GitHub star 数和最近推送时间\nGITHUB_TOKEN=<token> node scripts/update-open-source-stats.mjs\n```\n\nCI 使用 Node.js 24；本地需 Node.js 18+（使用了原生 `fetch` 和顶层 `await`）。\n\n## 架构与数据流\n\n### 单一数据源\n\n`resources/resources.json` 是所有资源的唯一数据源。README 中 `<!-- featured-resources:start -->` 和 `<!-- featured-resources:end -->` 之间的全部内容，以及 `<!-- resource-count:start/end -->` 徽章，**均由 `scripts/sync-readme.mjs` 自动覆盖**，手动编辑这些区块会在下次 CI 运行时被还原。\n\n### 脚本职责\n\n| 脚本 | 输入 | 输出 |\n|---|---|---|\n| `sync-readme.mjs` | `resources/resources.json`，`reports/availability.json` | `README.md`（精选区块） |\n| `check-availability.mjs` | `resources/resources.json`，`README.md` | `reports/availability.json`，`README.md`（状态标记） |\n| `validate-data.mjs` | 全部 JSON 文件 | 仅 stdout 错误报告 |\n| `update-open-source-stats.mjs` | `resources/resources.json`（GitHub API） | `resources/resources.json` |\n\n### 自动化流程\n\n- **每日 09:00 (北京时间)**：`check-availability.yml` 依次执行 sync-readme → check-availability → validate-data，提交结果\n- **PR / push to main**：`validate-data.yml` 校验数据格式\n- **Issue 审核**：维护者在资源推荐 Issue 下评论 `ok` 触发 `issue-to-resource-pr.yml`，自动将资源写入 resources.json 并发布\n\n## resources.json 数据约束\n\n新增或修改资源时必须满足以下规则（`validate-data.mjs` 会检测）：\n\n- `id`：小写字母和数字，以 `-` 分隔，模式 `^[a-z0-9]+(?:-[a-z0-9]+)*$`，全局唯一\n- `url`：必须是 HTTPS，`tvbox_config` 类型例外可用 HTTP\n- `category`：必须是枚举值之一：`online_video` / `video_app` / `cloud_search` / `magnet_search` / `subtitles` / `player` / `subscription` / `tvbox_config` / `membership` / `open_source` / `other`\n- `scores.*`：`more` / `speed` / `clean` / `stable` / `ease` 均为 1.0–5.0，最多一位小数\n- `risks.*`：`copyright` / `safety` / `privacy` / `payment` 枚举值：`low` / `medium` / `high` / `unknown`\n- `verification.status`：`pending` / `verified` / `recommended` / `caution` / `temporarily_unavailable` / `removed`\n- `featured: true` 的资源必须提供 `summary_short`（40 个字符以内）\n- `category: \"open_source\"` 的资源必须包含 `github.full_name`（`owner/repo` 格式）、`github.stars`（非负整数）、`github.pushed_at`（ISO 时间戳）\n- 日期字段格式：`YYYY-MM-DD`\n\n修改 resources.json 后，手动运行 `node scripts/sync-readme.mjs` 可在本地预览 README 变化。\n</yakaai_skills>"}