[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-yusufkaraaslan--Skill_Seekers":3,"tool-yusufkaraaslan--Skill_Seekers":64},[4,17,27,35,48,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,43,44,45,15,46,26,13,47],"数据工具","视频","插件","其他","音频",{"id":49,"name":50,"github_repo":51,"description_zh":52,"stars":53,"difficulty_score":10,"last_commit_at":54,"category_tags":55,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,46],{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},519,"PaddleOCR","PaddlePaddle\u002FPaddleOCR","PaddleOCR 是一款基于百度飞桨框架开发的高性能开源光学字符识别工具包。它的核心能力是将图片、PDF 等文档中的文字提取出来，转换成计算机可读取的结构化数据，让机器真正“看懂”图文内容。\n\n面对海量纸质或电子文档，PaddleOCR 解决了人工录入效率低、数字化成本高的问题。尤其在人工智能领域，它扮演着连接图像与大型语言模型（LLM）的桥梁角色，能将视觉信息直接转化为文本输入，助力智能问答、文档分析等应用场景落地。\n\nPaddleOCR 适合开发者、算法研究人员以及有文档自动化需求的普通用户。其技术优势十分明显：不仅支持全球 100 多种语言的识别，还能在 Windows、Linux、macOS 等多个系统上运行，并灵活适配 CPU、GPU、NPU 等各类硬件。作为一个轻量级且社区活跃的开源项目，PaddleOCR 既能满足快速集成的需求，也能支撑前沿的视觉语言研究，是处理文字识别任务的理想选择。",74913,"2026-04-05T10:44:17",[26,14,13,46],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":70,"readme_en":71,"readme_zh":72,"quickstart_zh":73,"use_case_zh":74,"hero_image_url":75,"owner_login":76,"owner_name":77,"owner_avatar_url":78,"owner_bio":79,"owner_company":79,"owner_location":79,"owner_email":80,"owner_twitter":79,"owner_website":79,"owner_url":81,"languages":82,"stars":98,"forks":99,"last_commit_at":100,"license":101,"difficulty_score":23,"env_os":102,"env_gpu":103,"env_ram":104,"env_deps":105,"category_tags":111,"github_topics":112,"view_count":131,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":132,"updated_at":133,"faqs":134,"releases":162},2578,"yusufkaraaslan\u002FSkill_Seekers","Skill_Seekers","Convert documentation websites, GitHub repositories, and PDFs into Claude AI skills with automatic conflict detection","Skill_Seekers 是一款专为 AI 系统打造的数据层工具，旨在将分散的知识源高效转化为结构化技能资产。它能自动抓取并处理文档网站、GitHub 仓库、PDF 文件、视频教程及笔记等多种格式的内容，将其快速封装为适用于 Claude、Gemini 等主流大模型的“技能包”，或直接接入 RAG（检索增强生成）流水线与 AI 编程助手。\n\n在开发自定义 AI 应用时，用户常面临数据源格式杂乱、内容冲突难以识别以及预处理耗时过长等痛点。Skill_Seekers 通过内置的自动冲突检测机制，智能识别并解决不同来源间的信息矛盾，确保知识库的准确性与一致性，将原本需要数小时的手工整理工作缩短至几分钟。\n\n这款工具特别适合开发者、AI 研究人员及技术团队使用。无论是希望为项目构建专属知识库的工程师，还是试图让 AI 助手深入理解特定技术栈的创作者，都能从中获益。其核心亮点在于支持超过 10 种数据源类型，提供丰富的预设配置，并原生集成 MCP（模型上下文协议），能够无缝嵌入现有的 LangChain、LlamaIndex 等工作流中。作为一个开源项目，Skill_Seekers 以简洁的命","Skill_Seekers 是一款专为 AI 系统打造的数据层工具，旨在将分散的知识源高效转化为结构化技能资产。它能自动抓取并处理文档网站、GitHub 仓库、PDF 文件、视频教程及笔记等多种格式的内容，将其快速封装为适用于 Claude、Gemini 等主流大模型的“技能包”，或直接接入 RAG（检索增强生成）流水线与 AI 编程助手。\n\n在开发自定义 AI 应用时，用户常面临数据源格式杂乱、内容冲突难以识别以及预处理耗时过长等痛点。Skill_Seekers 通过内置的自动冲突检测机制，智能识别并解决不同来源间的信息矛盾，确保知识库的准确性与一致性，将原本需要数小时的手工整理工作缩短至几分钟。\n\n这款工具特别适合开发者、AI 研究人员及技术团队使用。无论是希望为项目构建专属知识库的工程师，还是试图让 AI 助手深入理解特定技术栈的创作者，都能从中获益。其核心亮点在于支持超过 10 种数据源类型，提供丰富的预设配置，并原生集成 MCP（模型上下文协议），能够无缝嵌入现有的 LangChain、LlamaIndex 等工作流中。作为一个开源项目，Skill_Seekers 以简洁的命令行界面和灵活的扩展性，帮助用户轻松打通从原始数据到智能应用的“最后一公里”。","\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_86f7712e0c1b.png\" alt=\"Skill Seekers\" width=\"200\"\u002F>\n\u003C\u002Fp>\n\n# Skill Seekers\n\nEnglish | [简体中文](README.zh-CN.md) | [日本語](README.ja.md) | [한국어](README.ko.md) | [Español](README.es.md) | [Français](README.fr.md) | [Deutsch](README.de.md) | [Português](README.pt-BR.md) | [Türkçe](README.tr.md) | [العربية](README.ar.md) | [हिन्दी](README.hi.md) | [Русский](README.ru.md)\n\n[![Version](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fversion-3.5.0-blue.svg)](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Freleases)\n[![License: MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-yellow.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n[![Python 3.10+](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.10+-blue.svg)](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002F)\n[![MCP Integration](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMCP-Integrated-blue.svg)](https:\u002F\u002Fmodelcontextprotocol.io)\n[![Tested](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTests-2540%2B%20Passing-brightgreen.svg)](tests\u002F)\n[![Project Board](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FProject-Board-purple.svg)](https:\u002F\u002Fgithub.com\u002Fusers\u002Fyusufkaraaslan\u002Fprojects\u002F2)\n[![PyPI version](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Fskill-seekers.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F)\n[![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fskill-seekers.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F)\n[![PyPI - Python Version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fskill-seekers.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F)\n[![Website](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWebsite-skillseekersweb.com-blue.svg)](https:\u002F\u002Fskillseekersweb.com\u002F)\n[![Twitter Follow](https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Ffollow\u002F_yUSyUS_?style=social)](https:\u002F\u002Fx.com\u002F_yUSyUS_)\n[![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fyusufkaraaslan\u002FSkill_Seekers?style=social)](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers)\n[![PyPI Downloads](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_0cc707d5ff49.png)](https:\u002F\u002Fpepy.tech\u002Fprojects\u002Fskill-seekers)\n\n\u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F18329\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_1145cd82417e.png\" alt=\"yusufkaraaslan%2FSkill_Seekers | Trendshift\" style=\"width: 250px; height: 55px;\" width=\"250\" height=\"55\"\u002F>\u003C\u002Fa>\n\n**🧠 The data layer for AI systems.** Skill Seekers turns documentation sites, GitHub repos, PDFs, videos, notebooks, wikis, and 10+ more source types into structured knowledge assets—ready to power AI Skills (Claude, Gemini, OpenAI), RAG pipelines (LangChain, LlamaIndex, Pinecone), and AI coding assistants (Cursor, Windsurf, Cline) in minutes, not hours.\n\n> 🌐 **[Visit SkillSeekersWeb.com](https:\u002F\u002Fskillseekersweb.com\u002F)** - Browse 24+ preset configs, share your configs, and access complete documentation!\n\n> 📋 **[View Development Roadmap & Tasks](https:\u002F\u002Fgithub.com\u002Fusers\u002Fyusufkaraaslan\u002Fprojects\u002F2)** - 134 tasks across 10 categories, pick any to contribute!\n\n## 🌐 Ecosystem\n\nSkill Seekers is a multi-repo project. Here's where everything lives:\n\n| Repository | Description | Links |\n|-----------|-------------|-------|\n| **[Skill_Seekers](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers)** | Core CLI & MCP server (this repo) | [PyPI](https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F) |\n| **[skillseekersweb](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskillseekersweb)** | Website & documentation | [Live](https:\u002F\u002Fskillseekersweb.com\u002F) |\n| **[skill-seekers-configs](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskill-seekers-configs)** | Community config repository | |\n| **[skill-seekers-action](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskill-seekers-action)** | GitHub Action for CI\u002FCD | |\n| **[skill-seekers-plugin](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskill-seekers-plugin)** | Claude Code plugin | |\n| **[homebrew-skill-seekers](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fhomebrew-skill-seekers)** | Homebrew tap for macOS | |\n\n> **Want to contribute?** The website and configs repos are great starting points for new contributors!\n\n## 🧠 The Data Layer for AI Systems\n\n**Skill Seekers is the universal preprocessing layer** that sits between raw documentation and every AI system that consumes it. Whether you are building Claude skills, a LangChain RAG pipeline, or a Cursor `.cursorrules` file — the data preparation is identical. You do it once, and export to all targets.\n\n```bash\n# One command → structured knowledge asset\nskill-seekers create https:\u002F\u002Fdocs.react.dev\u002F\n# or: skill-seekers create facebook\u002Freact\n# or: skill-seekers create .\u002Fmy-project\n\n# Export to any AI system\nskill-seekers package output\u002Freact --target claude      # → Claude AI Skill (ZIP)\nskill-seekers package output\u002Freact --target langchain   # → LangChain Documents\nskill-seekers package output\u002Freact --target llama-index # → LlamaIndex TextNodes\nskill-seekers package output\u002Freact --target cursor      # → .cursorrules\n```\n\n### What gets built\n\n| Output | Target | What it powers |\n|--------|--------|---------------|\n| **Claude Skill** (ZIP + YAML) | `--target claude` | Claude Code, Claude API |\n| **Gemini Skill** (tar.gz) | `--target gemini` | Google Gemini |\n| **OpenAI \u002F Custom GPT** (ZIP) | `--target openai` | GPT-4o, custom assistants |\n| **LangChain Documents** | `--target langchain` | QA chains, agents, retrievers |\n| **LlamaIndex TextNodes** | `--target llama-index` | Query engines, chat engines |\n| **Haystack Documents** | `--target haystack` | Enterprise RAG pipelines |\n| **Pinecone-ready** (Markdown) | `--target markdown` | Vector upsert |\n| **ChromaDB \u002F FAISS \u002F Qdrant** | `--format chroma\u002Ffaiss\u002Fqdrant` | Local vector DBs |\n| **Cursor** `.cursorrules` | `--target claude` → copy | Cursor IDE AI context |\n| **Windsurf \u002F Cline \u002F Continue** | `--target claude` → copy | VS Code, IntelliJ, Vim |\n\n### Why it matters\n\n- ⚡ **99% faster** — Days of manual data prep → 15–45 minutes\n- 🎯 **AI Skill quality** — 500+ line SKILL.md files with examples, patterns, and guides\n- 📊 **RAG-ready chunks** — Smart chunking preserves code blocks and maintains context\n- 🎬 **Videos** — Extract code, transcripts, and structured knowledge from YouTube and local videos\n- 🔄 **Multi-source** — Combine 17 source types (docs, GitHub, PDFs, videos, notebooks, wikis, and more) into one knowledge asset\n- 🌐 **One prep, every target** — Export the same asset to 16 platforms without re-scraping\n- ✅ **Battle-tested** — 2,540+ tests, 24+ framework presets, production-ready\n\n## 🚀 Quick Start (3 Commands)\n\n```bash\n# 1. Install\npip install skill-seekers\n\n# 2. Create skill from any source\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F\n\n# 3. Package for your AI platform\nskill-seekers package output\u002Fdjango --target claude\n```\n\n**That's it!** You now have `output\u002Fdjango-claude.zip` ready to use.\n\n```bash\n# Use a different AI agent for enhancement (default: claude)\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F --agent kimi\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F --agent codex\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F --agent-cmd \"my-custom-agent run\"\n```\n\n### Other Sources (17 Supported)\n\n```bash\n# GitHub repository\nskill-seekers create facebook\u002Freact\n\n# Local project\nskill-seekers create .\u002Fmy-project\n\n# PDF document\nskill-seekers create manual.pdf\n\n# Word document\nskill-seekers create report.docx\n\n# EPUB e-book\nskill-seekers create book.epub\n\n# Jupyter Notebook\nskill-seekers create notebook.ipynb\n\n# OpenAPI spec\nskill-seekers create openapi.yaml\n\n# PowerPoint presentation\nskill-seekers create presentation.pptx\n\n# AsciiDoc document\nskill-seekers create guide.adoc\n\n# Local HTML file\nskill-seekers create page.html\n\n# RSS\u002FAtom feed\nskill-seekers create feed.rss\n\n# Man page\nskill-seekers create curl.1\n\n# Video (YouTube, Vimeo, or local file — requires skill-seekers[video])\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --name mytutorial\n# First time? Auto-install GPU-aware visual deps:\nskill-seekers video --setup\n\n# Confluence wiki\nskill-seekers confluence --space TEAM --name wiki\n\n# Notion pages\nskill-seekers notion --database-id ... --name docs\n\n# Slack\u002FDiscord chat export\nskill-seekers chat --export-dir .\u002Fslack-export --name team-chat\n```\n\n### Export Everywhere\n\n```bash\n# Package for multiple platforms\nfor platform in claude gemini openai langchain; do\n  skill-seekers package output\u002Fdjango --target $platform\ndone\n```\n\n## What is Skill Seekers?\n\nSkill Seekers is the **data layer for AI systems**. It transforms 17 source types—documentation websites, GitHub repositories, PDFs, videos, Jupyter Notebooks, Word\u002FEPUB\u002FAsciiDoc documents, OpenAPI specs, PowerPoint presentations, RSS feeds, man pages, Confluence wikis, Notion pages, Slack\u002FDiscord exports, and more—into structured knowledge assets for every AI target:\n\n| Use Case | What you get | Examples |\n|----------|-------------|---------|\n| **AI Skills** | Comprehensive SKILL.md + references | Claude Code, Gemini, GPT |\n| **RAG Pipelines** | Chunked documents with rich metadata | LangChain, LlamaIndex, Haystack |\n| **Vector Databases** | Pre-formatted data ready for upsert | Pinecone, Chroma, Weaviate, FAISS |\n| **AI Coding Assistants** | Context files your IDE AI reads automatically | Cursor, Windsurf, Cline, Continue.dev |\n\n## 📚 Documentation\n\n| I want to... | Read this |\n|--------------|-----------|\n| **Get started quickly** | [Quick Start](docs\u002Fgetting-started\u002F02-quick-start.md) - 3 commands to first skill |\n| **Understand concepts** | [Core Concepts](docs\u002Fuser-guide\u002F01-core-concepts.md) - How it works |\n| **Scrape sources** | [Scraping Guide](docs\u002Fuser-guide\u002F02-scraping.md) - All source types |\n| **Enhance skills** | [Enhancement Guide](docs\u002Fuser-guide\u002F03-enhancement.md) - AI enhancement |\n| **Export skills** | [Packaging Guide](docs\u002Fuser-guide\u002F04-packaging.md) - Platform export |\n| **Look up commands** | [CLI Reference](docs\u002Freference\u002FCLI_REFERENCE.md) - All 20 commands |\n| **Configure** | [Config Format](docs\u002Freference\u002FCONFIG_FORMAT.md) - JSON specification |\n| **Fix issues** | [Troubleshooting](docs\u002Fuser-guide\u002F06-troubleshooting.md) - Common problems |\n\n**Complete documentation:** [docs\u002FREADME.md](docs\u002FREADME.md)\n\nInstead of spending days on manual preprocessing, Skill Seekers:\n\n1. **Ingests** — docs, GitHub repos, local codebases, PDFs, videos, notebooks, wikis, and 10+ more source types\n2. **Analyzes** — deep AST parsing, pattern detection, API extraction\n3. **Structures** — categorized reference files with metadata\n4. **Enhances** — AI-powered SKILL.md generation (Claude, Gemini, or local)\n5. **Exports** — 16 platform-specific formats from one asset\n\n## Why Use This?\n\n### For AI Skill Builders (Claude, Gemini, OpenAI)\n\n- 🎯 **Production-grade Skills** — 500+ line SKILL.md files with code examples, patterns, and guides\n- 🔄 **Enhancement Workflows** — Apply `security-focus`, `architecture-comprehensive`, or custom YAML presets\n- 🎮 **Any Domain** — Game engines (Godot, Unity), frameworks (React, Django), internal tools\n- 🔧 **Teams** — Combine internal docs + code into a single source of truth\n- 📚 **Quality** — AI-enhanced with examples, quick reference, and navigation guidance\n\n### For RAG Builders & AI Engineers\n\n- 🤖 **RAG-ready data** — Pre-chunked LangChain `Documents`, LlamaIndex `TextNodes`, Haystack `Documents`\n- 🚀 **99% faster** — Days of preprocessing → 15–45 minutes\n- 📊 **Smart metadata** — Categories, sources, types → better retrieval accuracy\n- 🔄 **Multi-source** — Combine docs + GitHub + PDFs + videos in one pipeline\n- 🌐 **Platform-agnostic** — Export to any vector DB or framework without re-scraping\n\n### For AI Coding Assistant Users\n\n- 💻 **Cursor \u002F Windsurf \u002F Cline** — Generate `.cursorrules` \u002F `.windsurfrules` \u002F `.clinerules` automatically\n- 🎯 **Persistent context** — AI \"knows\" your frameworks without repeated prompting\n- 📚 **Always current** — Update context in minutes when docs change\n\n## Key Features\n\n### 🌐 Documentation Scraping\n- ✅ **Smart SPA Discovery** - Three-layer discovery for JavaScript SPA sites (sitemap.xml → llms.txt → headless browser rendering)\n- ✅ **llms.txt Support** - Automatically detects and uses LLM-ready documentation files (10x faster)\n- ✅ **Universal Scraper** - Works with ANY documentation website\n- ✅ **Smart Categorization** - Automatically organizes content by topic\n- ✅ **Code Language Detection** - Recognizes Python, JavaScript, C++, GDScript, etc.\n- ✅ **24+ Ready-to-Use Presets** - Godot, React, Vue, Django, FastAPI, and more\n\n### 📄 PDF Support\n- ✅ **Basic PDF Extraction** - Extract text, code, and images from PDF files\n- ✅ **OCR for Scanned PDFs** - Extract text from scanned documents\n- ✅ **Password-Protected PDFs** - Handle encrypted PDFs\n- ✅ **Table Extraction** - Extract complex tables from PDFs\n- ✅ **Parallel Processing** - 3x faster for large PDFs\n- ✅ **Intelligent Caching** - 50% faster on re-runs\n\n### 🎬 Video Extraction\n- ✅ **YouTube & Local Videos** - Extract transcripts, on-screen code, and structured knowledge from videos\n- ✅ **Visual Frame Analysis** - OCR extraction from code editors, terminals, slides, and diagrams\n- ✅ **GPU Auto-Detection** - Automatically installs correct PyTorch build (CUDA\u002FROCm\u002FMPS\u002FCPU)\n- ✅ **AI Enhancement** - Two-pass: clean OCR artifacts + generate polished SKILL.md\n- ✅ **Time Clipping** - Extract specific sections with `--start-time` and `--end-time`\n- ✅ **Playlist Support** - Batch process all videos in a YouTube playlist\n- ✅ **Vision API Fallback** - Use Claude Vision for low-confidence OCR frames\n\n### 🐙 GitHub Repository Analysis\n- ✅ **Deep Code Analysis** - AST parsing for Python, JavaScript, TypeScript, Java, C++, Go\n- ✅ **API Extraction** - Functions, classes, methods with parameters and types\n- ✅ **Repository Metadata** - README, file tree, language breakdown, stars\u002Fforks\n- ✅ **GitHub Issues & PRs** - Fetch open\u002Fclosed issues with labels and milestones\n- ✅ **CHANGELOG & Releases** - Automatically extract version history\n- ✅ **Conflict Detection** - Compare documented APIs vs actual code implementation\n- ✅ **MCP Integration** - Natural language: \"Scrape GitHub repo facebook\u002Freact\"\n\n### 🔄 Unified Multi-Source Scraping\n- ✅ **Combine Multiple Sources** - Mix documentation + GitHub + PDF in one skill\n- ✅ **Conflict Detection** - Automatically finds discrepancies between docs and code\n- ✅ **Intelligent Merging** - Rule-based or AI-powered conflict resolution\n- ✅ **Transparent Reporting** - Side-by-side comparison with ⚠️ warnings\n- ✅ **Documentation Gap Analysis** - Identifies outdated docs and undocumented features\n- ✅ **Single Source of Truth** - One skill showing both intent (docs) and reality (code)\n- ✅ **Backward Compatible** - Legacy single-source configs still work\n\n### 🤖 Multi-LLM Platform Support\n- ✅ **12 LLM Platforms** - Claude AI, Google Gemini, OpenAI ChatGPT, MiniMax AI, Generic Markdown, OpenCode, Kimi (Moonshot AI), DeepSeek AI, Qwen (Alibaba), OpenRouter, Together AI, Fireworks AI\n- ✅ **Universal Scraping** - Same documentation works for all platforms\n- ✅ **Platform-Specific Packaging** - Optimized formats for each LLM\n- ✅ **One-Command Export** - `--target` flag selects platform\n- ✅ **Optional Dependencies** - Install only what you need\n- ✅ **100% Backward Compatible** - Existing Claude workflows unchanged\n\n| Platform | Format | Upload | Enhancement | API Key | Custom Endpoint |\n|----------|--------|--------|-------------|---------|-----------------|\n| **Claude AI** | ZIP + YAML | ✅ Auto | ✅ Yes | ANTHROPIC_API_KEY | ANTHROPIC_BASE_URL |\n| **Google Gemini** | tar.gz | ✅ Auto | ✅ Yes | GOOGLE_API_KEY | - |\n| **OpenAI ChatGPT** | ZIP + Vector Store | ✅ Auto | ✅ Yes | OPENAI_API_KEY | - |\n| **MiniMax AI** | ZIP + Knowledge Files | ✅ Auto | ✅ Yes | MINIMAX_API_KEY | - |\n| **Generic Markdown** | ZIP | ❌ Manual | ❌ No | - | - |\n\n```bash\n# Claude (default - no changes needed!)\nskill-seekers package output\u002Freact\u002F\nskill-seekers upload react.zip\n\n# Google Gemini\npip install skill-seekers[gemini]\nskill-seekers package output\u002Freact\u002F --target gemini\nskill-seekers upload react-gemini.tar.gz --target gemini\n\n# OpenAI ChatGPT\npip install skill-seekers[openai]\nskill-seekers package output\u002Freact\u002F --target openai\nskill-seekers upload react-openai.zip --target openai\n\n# MiniMax AI\npip install skill-seekers[minimax]\nskill-seekers package output\u002Freact\u002F --target minimax\nskill-seekers upload react-minimax.zip --target minimax\n\n# Generic Markdown (universal export)\nskill-seekers package output\u002Freact\u002F --target markdown\n# Use the markdown files directly in any LLM\n```\n\n\u003Cdetails>\n\u003Csummary>🔧 \u003Cstrong>Environment Variables for Claude-Compatible APIs (e.g., GLM-4.7)\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nSkill Seekers supports any Claude-compatible API endpoint:\n\n```bash\n# Option 1: Official Anthropic API (default)\nexport ANTHROPIC_API_KEY=sk-ant-...\n\n# Option 2: GLM-4.7 Claude-compatible API\nexport ANTHROPIC_API_KEY=your-glm-47-api-key\nexport ANTHROPIC_BASE_URL=https:\u002F\u002Fglm-4-7-endpoint.com\u002Fv1\n\n# All AI enhancement features will use the configured endpoint\nskill-seekers enhance output\u002Freact\u002F\nskill-seekers analyze --directory . --enhance\n```\n\n**Note**: Setting `ANTHROPIC_BASE_URL` allows you to use any Claude-compatible API endpoint, such as GLM-4.7 (智谱 AI) or other compatible services.\n\n\u003C\u002Fdetails>\n\n**Installation:**\n```bash\n# Install with Gemini support\npip install skill-seekers[gemini]\n\n# Install with OpenAI support\npip install skill-seekers[openai]\n\n# Install with MiniMax support\npip install skill-seekers[minimax]\n\n# Install with all LLM platforms\npip install skill-seekers[all-llms]\n```\n\n### 🔗 RAG Framework Integrations\n\n- ✅ **LangChain Documents** - Direct export to `Document` format with `page_content` + metadata\n  - Perfect for: QA chains, retrievers, vector stores, agents\n  - Example: [LangChain RAG Pipeline](examples\u002Flangchain-rag-pipeline\u002F)\n  - Guide: [LangChain Integration](docs\u002Fintegrations\u002FLANGCHAIN.md)\n\n- ✅ **LlamaIndex TextNodes** - Export to `TextNode` format with unique IDs + embeddings\n  - Perfect for: Query engines, chat engines, storage context\n  - Example: [LlamaIndex Query Engine](examples\u002Fllama-index-query-engine\u002F)\n  - Guide: [LlamaIndex Integration](docs\u002Fintegrations\u002FLLAMA_INDEX.md)\n\n- ✅ **Pinecone-Ready Format** - Optimized for vector database upsert\n  - Perfect for: Production vector search, semantic search, hybrid search\n  - Example: [Pinecone Upsert](examples\u002Fpinecone-upsert\u002F)\n  - Guide: [Pinecone Integration](docs\u002Fintegrations\u002FPINECONE.md)\n\n**Quick Export:**\n```bash\n# LangChain Documents (JSON)\nskill-seekers package output\u002Fdjango --target langchain\n# → output\u002Fdjango-langchain.json\n\n# LlamaIndex TextNodes (JSON)\nskill-seekers package output\u002Fdjango --target llama-index\n# → output\u002Fdjango-llama-index.json\n\n# Markdown (Universal)\nskill-seekers package output\u002Fdjango --target markdown\n# → output\u002Fdjango-markdown\u002FSKILL.md + references\u002F\n```\n\n**Complete RAG Pipeline Guide:** [RAG Pipelines Documentation](docs\u002Fintegrations\u002FRAG_PIPELINES.md)\n\n---\n\n### 🧠 AI Coding Assistant Integrations\n\nTransform any framework documentation into expert coding context for 4+ AI assistants:\n\n- ✅ **Cursor IDE** - Generate `.cursorrules` for AI-powered code suggestions\n  - Perfect for: Framework-specific code generation, consistent patterns\n  - Works with: Cursor IDE (VS Code fork)\n  - Guide: [Cursor Integration](docs\u002Fintegrations\u002FCURSOR.md)\n  - Example: [Cursor React Skill](examples\u002Fcursor-react-skill\u002F)\n\n- ✅ **Windsurf** - Customize Windsurf's AI assistant context with `.windsurfrules`\n  - Perfect for: IDE-native AI assistance, flow-based coding\n  - Works with: Windsurf IDE by Codeium\n  - Guide: [Windsurf Integration](docs\u002Fintegrations\u002FWINDSURF.md)\n  - Example: [Windsurf FastAPI Context](examples\u002Fwindsurf-fastapi-context\u002F)\n\n- ✅ **Cline (VS Code)** - System prompts + MCP for VS Code agent\n  - Perfect for: Agentic code generation in VS Code\n  - Works with: Cline extension for VS Code\n  - Guide: [Cline Integration](docs\u002Fintegrations\u002FCLINE.md)\n  - Example: [Cline Django Assistant](examples\u002Fcline-django-assistant\u002F)\n\n- ✅ **Continue.dev** - Context servers for IDE-agnostic AI\n  - Perfect for: Multi-IDE environments (VS Code, JetBrains, Vim), custom LLM providers\n  - Works with: Any IDE with Continue.dev plugin\n  - Guide: [Continue Integration](docs\u002Fintegrations\u002FCONTINUE_DEV.md)\n  - Example: [Continue Universal Context](examples\u002Fcontinue-dev-universal\u002F)\n\n**Quick Export for AI Coding Tools:**\n```bash\n# For any AI coding assistant (Cursor, Windsurf, Cline, Continue.dev)\nskill-seekers scrape --config configs\u002Fdjango.json\nskill-seekers package output\u002Fdjango --target claude  # or --target markdown\n\n# Copy to your project (example for Cursor)\ncp output\u002Fdjango-claude\u002FSKILL.md my-project\u002F.cursorrules\n\n# Or for Windsurf\ncp output\u002Fdjango-claude\u002FSKILL.md my-project\u002F.windsurf\u002Frules\u002Fdjango.md\n\n# Or for Cline\ncp output\u002Fdjango-claude\u002FSKILL.md my-project\u002F.clinerules\n\n# Or for Continue.dev (HTTP server)\npython examples\u002Fcontinue-dev-universal\u002Fcontext_server.py\n# Configure in ~\u002F.continue\u002Fconfig.json\n```\n\n**Integration Hub:** [All AI System Integrations](docs\u002Fintegrations\u002FINTEGRATIONS.md)\n\n---\n\n### 🌊 Three-Stream GitHub Architecture\n- ✅ **Triple-Stream Analysis** - Split GitHub repos into Code, Docs, and Insights streams\n- ✅ **Unified Codebase Analyzer** - Works with GitHub URLs AND local paths\n- ✅ **C3.x as Analysis Depth** - Choose 'basic' (1-2 min) or 'c3x' (20-60 min) analysis\n- ✅ **Enhanced Router Generation** - GitHub metadata, README quick start, common issues\n- ✅ **Issue Integration** - Top problems and solutions from GitHub issues\n- ✅ **Smart Routing Keywords** - GitHub labels weighted 2x for better topic detection\n\n**Three Streams Explained:**\n- **Stream 1: Code** - Deep C3.x analysis (patterns, examples, guides, configs, architecture)\n- **Stream 2: Docs** - Repository documentation (README, CONTRIBUTING, docs\u002F*.md)\n- **Stream 3: Insights** - Community knowledge (issues, labels, stars, forks)\n\n```python\nfrom skill_seekers.cli.unified_codebase_analyzer import UnifiedCodebaseAnalyzer\n\n# Analyze GitHub repo with all three streams\nanalyzer = UnifiedCodebaseAnalyzer()\nresult = analyzer.analyze(\n    source=\"https:\u002F\u002Fgithub.com\u002Ffacebook\u002Freact\",\n    depth=\"c3x\",  # or \"basic\" for fast analysis\n    fetch_github_metadata=True\n)\n\n# Access code stream (C3.x analysis)\nprint(f\"Design patterns: {len(result.code_analysis['c3_1_patterns'])}\")\nprint(f\"Test examples: {result.code_analysis['c3_2_examples_count']}\")\n\n# Access docs stream (repository docs)\nprint(f\"README: {result.github_docs['readme'][:100]}\")\n\n# Access insights stream (GitHub metadata)\nprint(f\"Stars: {result.github_insights['metadata']['stars']}\")\nprint(f\"Common issues: {len(result.github_insights['common_problems'])}\")\n```\n\n**See complete documentation**: [Three-Stream Implementation Summary](docs\u002FIMPLEMENTATION_SUMMARY_THREE_STREAM.md)\n\n### 🔐 Smart Rate Limit Management & Configuration\n- ✅ **Multi-Token Configuration System** - Manage multiple GitHub accounts (personal, work, OSS)\n  - Secure config storage at `~\u002F.config\u002Fskill-seekers\u002Fconfig.json` (600 permissions)\n  - Per-profile rate limit strategies: `prompt`, `wait`, `switch`, `fail`\n  - Configurable timeout per profile (default: 30 min, prevents indefinite waits)\n  - Smart fallback chain: CLI arg → Env var → Config file → Prompt\n  - API key management for Claude, Gemini, OpenAI\n- ✅ **Interactive Configuration Wizard** - Beautiful terminal UI for easy setup\n  - Browser integration for token creation (auto-opens GitHub, etc.)\n  - Token validation and connection testing\n  - Visual status display with color coding\n- ✅ **Intelligent Rate Limit Handler** - No more indefinite waits!\n  - Upfront warning about rate limits (60\u002Fhour vs 5000\u002Fhour)\n  - Real-time detection from GitHub API responses\n  - Live countdown timers with progress\n  - Automatic profile switching when rate limited\n  - Four strategies: prompt (ask), wait (countdown), switch (try another), fail (abort)\n- ✅ **Resume Capability** - Continue interrupted jobs\n  - Auto-save progress at configurable intervals (default: 60 sec)\n  - List all resumable jobs with progress details\n  - Auto-cleanup of old jobs (default: 7 days)\n- ✅ **CI\u002FCD Support** - Non-interactive mode for automation\n  - `--non-interactive` flag fails fast without prompts\n  - `--profile` flag to select specific GitHub account\n  - Clear error messages for pipeline logs\n\n**Quick Setup:**\n```bash\n# One-time configuration (5 minutes)\nskill-seekers config --github\n\n# Use specific profile for private repos\nskill-seekers github --repo mycompany\u002Fprivate-repo --profile work\n\n# CI\u002FCD mode (fail fast, no prompts)\nskill-seekers github --repo owner\u002Frepo --non-interactive\n\n# Resume interrupted job\nskill-seekers resume --list\nskill-seekers resume github_react_20260117_143022\n```\n\n**Rate Limit Strategies Explained:**\n- **prompt** (default) - Ask what to do when rate limited (wait, switch, setup token, cancel)\n- **wait** - Automatically wait with countdown timer (respects timeout)\n- **switch** - Automatically try next available profile (for multi-account setups)\n- **fail** - Fail immediately with clear error (perfect for CI\u002FCD)\n\n### 🎯 Bootstrap Skill - Self-Hosting\n\nGenerate skill-seekers as a skill to use within your AI agent (Claude Code, Kimi, Codex, etc.):\n\n```bash\n# Generate the skill\n.\u002Fscripts\u002Fbootstrap_skill.sh\n\n# Install to Claude Code\ncp -r output\u002Fskill-seekers ~\u002F.claude\u002Fskills\u002F\n```\n\n**What you get:**\n- ✅ **Complete skill documentation** - All CLI commands and usage patterns\n- ✅ **CLI command reference** - Every tool and its options documented\n- ✅ **Quick start examples** - Common workflows and best practices\n- ✅ **Auto-generated API docs** - Code analysis, patterns, and examples\n\n### 🔐 Private Config Repositories\n- ✅ **Git-Based Config Sources** - Fetch configs from private\u002Fteam git repositories\n- ✅ **Multi-Source Management** - Register unlimited GitHub, GitLab, Bitbucket repos\n- ✅ **Team Collaboration** - Share custom configs across 3-5 person teams\n- ✅ **Enterprise Support** - Scale to 500+ developers with priority-based resolution\n- ✅ **Secure Authentication** - Environment variable tokens (GITHUB_TOKEN, GITLAB_TOKEN)\n- ✅ **Intelligent Caching** - Clone once, pull updates automatically\n- ✅ **Offline Mode** - Work with cached configs when offline\n\n### 🤖 Codebase Analysis (C3.x)\n\n**C3.4: Configuration Pattern Extraction with AI Enhancement**\n- ✅ **9 Config Formats** - JSON, YAML, TOML, ENV, INI, Python, JavaScript, Dockerfile, Docker Compose\n- ✅ **7 Pattern Types** - Database, API, logging, cache, email, auth, server configurations\n- ✅ **AI Enhancement** - Optional dual-mode AI analysis (API + LOCAL)\n  - Explains what each config does\n  - Suggests best practices and improvements\n  - **Security analysis** - Finds hardcoded secrets, exposed credentials\n- ✅ **Auto-Documentation** - Generates JSON + Markdown documentation of all configs\n- ✅ **MCP Integration** - `extract_config_patterns` tool with enhancement support\n\n**C3.3: AI-Enhanced How-To Guides**\n- ✅ **Comprehensive AI Enhancement** - Transforms basic guides into professional tutorials\n- ✅ **5 Automatic Improvements** - Step descriptions, troubleshooting, prerequisites, next steps, use cases\n- ✅ **Dual-Mode Support** - API mode (Claude API) or LOCAL mode (Claude Code CLI)\n- ✅ **No API Costs with LOCAL Mode** - FREE enhancement using your Claude Code Max plan\n- ✅ **Quality Transformation** - 75-line templates → 500+ line comprehensive guides\n\n**Usage:**\n```bash\n# Quick analysis (1-2 min, basic features only)\nskill-seekers analyze --directory tests\u002F --quick\n\n# Comprehensive analysis with AI (20-60 min, all features)\nskill-seekers analyze --directory tests\u002F --comprehensive\n\n# With AI enhancement\nskill-seekers analyze --directory tests\u002F --enhance\n```\n\n**Full Documentation:** [docs\u002FHOW_TO_GUIDES.md](docs\u002FHOW_TO_GUIDES.md#ai-enhancement-new)\n\n### 🔄 Enhancement Workflow Presets\n\nReusable YAML-defined enhancement pipelines that control how AI transforms your raw documentation into a polished skill.\n\n- ✅ **5 Bundled Presets** — `default`, `minimal`, `security-focus`, `architecture-comprehensive`, `api-documentation`\n- ✅ **User-Defined Presets** — add custom workflows to `~\u002F.config\u002Fskill-seekers\u002Fworkflows\u002F`\n- ✅ **Multiple Workflows** — chain two or more workflows in one command\n- ✅ **Fully Managed CLI** — list, inspect, copy, add, remove, and validate workflows\n\n```bash\n# Apply a single workflow\nskill-seekers create .\u002Fmy-project --enhance-workflow security-focus\n\n# Chain multiple workflows (applied in order)\nskill-seekers create .\u002Fmy-project \\\n  --enhance-workflow security-focus \\\n  --enhance-workflow minimal\n\n# Manage presets\nskill-seekers workflows list                          # List all (bundled + user)\nskill-seekers workflows show security-focus           # Print YAML content\nskill-seekers workflows copy security-focus           # Copy to user dir for editing\nskill-seekers workflows add .\u002Fmy-workflow.yaml        # Install a custom preset\nskill-seekers workflows remove my-workflow            # Remove a user preset\nskill-seekers workflows validate security-focus       # Validate preset structure\n\n# Copy multiple at once\nskill-seekers workflows copy security-focus minimal api-documentation\n\n# Add multiple files at once\nskill-seekers workflows add .\u002Fwf-a.yaml .\u002Fwf-b.yaml\n\n# Remove multiple at once\nskill-seekers workflows remove my-wf-a my-wf-b\n```\n\n**YAML preset format:**\n```yaml\nname: security-focus\ndescription: \"Security-focused review: vulnerabilities, auth, data handling\"\nversion: \"1.0\"\nstages:\n  - name: vulnerabilities\n    type: custom\n    prompt: \"Review for OWASP top 10 and common security vulnerabilities...\"\n  - name: auth-review\n    type: custom\n    prompt: \"Examine authentication and authorisation patterns...\"\n    uses_history: true\n```\n\n### ⚡ Performance & Scale\n- ✅ **Async Mode** - 2-3x faster scraping with async\u002Fawait (use `--async` flag)\n- ✅ **Large Documentation Support** - Handle 10K-40K+ page docs with intelligent splitting\n- ✅ **Router\u002FHub Skills** - Intelligent routing to specialized sub-skills\n- ✅ **Parallel Scraping** - Process multiple skills simultaneously\n- ✅ **Checkpoint\u002FResume** - Never lose progress on long scrapes\n- ✅ **Caching System** - Scrape once, rebuild instantly\n\n### 🤖 Agent-Agnostic Skill Generation\n- ✅ **Multi-Agent Support** - Generate skills for Claude, Kimi, Codex, Copilot, OpenCode, or any custom agent via `--agent` flag\n- ✅ **Custom Agent Commands** - Use `--agent-cmd` to specify a custom agent CLI command for enhancement\n- ✅ **Universal Flags** - `--agent` and `--agent-cmd` available on all commands (create, scrape, github, pdf, etc.)\n\n### 📦 Marketplace Pipeline\n- ✅ **Publish to Marketplace** - Publish skills to Claude Code plugin marketplace repos\n- ✅ **End-to-End Pipeline** - From documentation source to published marketplace entry\n\n### ✅ Quality Assurance\n- ✅ **Fully Tested** - 2,540+ tests with comprehensive coverage\n\n---\n\n## 📦 Installation\n\n```bash\n# Basic install (documentation scraping, GitHub analysis, PDF, packaging)\npip install skill-seekers\n\n# With all LLM platform support\npip install skill-seekers[all-llms]\n\n# With MCP server\npip install skill-seekers[mcp]\n\n# Everything\npip install skill-seekers[all]\n```\n\n**Need help choosing?** Run the setup wizard:\n```bash\nskill-seekers-setup\n```\n\n### Installation Options\n\n| Install | Features |\n|---------|----------|\n| `pip install skill-seekers` | Scraping, GitHub analysis, PDF, all platforms |\n| `pip install skill-seekers[gemini]` | + Google Gemini support |\n| `pip install skill-seekers[openai]` | + OpenAI ChatGPT support |\n| `pip install skill-seekers[all-llms]` | + All LLM platforms |\n| `pip install skill-seekers[mcp]` | + MCP server for Claude Code, Cursor, etc. |\n| `pip install skill-seekers[video]` | + YouTube\u002FVimeo transcript & metadata extraction |\n| `pip install skill-seekers[video-full]` | + Whisper transcription & visual frame extraction |\n| `pip install skill-seekers[jupyter]` | + Jupyter Notebook support |\n| `pip install skill-seekers[pptx]` | + PowerPoint support |\n| `pip install skill-seekers[confluence]` | + Confluence wiki support |\n| `pip install skill-seekers[notion]` | + Notion pages support |\n| `pip install skill-seekers[rss]` | + RSS\u002FAtom feed support |\n| `pip install skill-seekers[chat]` | + Slack\u002FDiscord chat export support |\n| `pip install skill-seekers[asciidoc]` | + AsciiDoc document support |\n| `pip install skill-seekers[all]` | Everything enabled |\n\n> **Video visual deps (GPU-aware):** After installing `skill-seekers[video-full]`, run\n> `skill-seekers video --setup` to auto-detect your GPU and install the correct PyTorch\n> variant + easyocr. This is the recommended way to install visual extraction dependencies.\n\n---\n\n## 🚀 One-Command Install Workflow\n\n**The fastest way to go from config to uploaded skill - complete automation:**\n\n```bash\n# Install React skill from official configs (auto-uploads to Claude)\nskill-seekers install --config react\n\n# Install from local config file\nskill-seekers install --config configs\u002Fcustom.json\n\n# Install without uploading (package only)\nskill-seekers install --config django --no-upload\n\n# Preview workflow without executing\nskill-seekers install --config react --dry-run\n```\n\n**Time:** 20-45 minutes total | **Quality:** Production-ready (9\u002F10) | **Cost:** Free\n\n**Phases executed:**\n```\n📥 PHASE 1: Fetch Config (if config name provided)\n📖 PHASE 2: Scrape Documentation\n✨ PHASE 3: AI Enhancement (MANDATORY - no skip option)\n📦 PHASE 4: Package Skill\n☁️  PHASE 5: Upload to Claude (optional, requires API key)\n```\n\n**Requirements:**\n- ANTHROPIC_API_KEY environment variable (for auto-upload)\n- Claude Code Max plan (for local AI enhancement), or use `--agent` to select a different AI agent\n\n---\n\n## 📊 Feature Matrix\n\nSkill Seekers supports **12 LLM platforms**, **17 source types**, and full feature parity across all targets.\n\n**Platforms:** Claude AI, Google Gemini, OpenAI ChatGPT, MiniMax AI, Generic Markdown, OpenCode, Kimi (Moonshot AI), DeepSeek AI, Qwen (Alibaba), OpenRouter, Together AI, Fireworks AI\n**Source Types:** Documentation websites, GitHub repos, PDFs, Word (.docx), EPUB, Video, Local codebases, Jupyter Notebooks, Local HTML, OpenAPI\u002FSwagger, AsciiDoc, PowerPoint (.pptx), RSS\u002FAtom feeds, Man pages, Confluence wikis, Notion pages, Slack\u002FDiscord chat exports\n\nSee [Complete Feature Matrix](docs\u002FFEATURE_MATRIX.md) for detailed platform and feature support.\n\n### Quick Platform Comparison\n\n| Feature | Claude | Gemini | OpenAI | MiniMax | Markdown |\n|---------|--------|--------|--------|--------|----------|\n| Format | ZIP + YAML | tar.gz | ZIP + Vector | ZIP + Knowledge | ZIP |\n| Upload | ✅ API | ✅ API | ✅ API | ✅ API | ❌ Manual |\n| Enhancement | ✅ Sonnet 4 | ✅ 2.0 Flash | ✅ GPT-4o | ✅ M2.7 | ❌ None |\n| All Skill Modes | ✅ | ✅ | ✅ | ✅ | ✅ |\n\n---\n\n## Usage Examples\n\n### Documentation Scraping\n\n```bash\n# Scrape documentation website\nskill-seekers scrape --config configs\u002Freact.json\n\n# Quick scrape without config\nskill-seekers scrape --url https:\u002F\u002Freact.dev --name react\n\n# With async mode (3x faster)\nskill-seekers scrape --config configs\u002Fgodot.json --async --workers 8\n\n# Use a specific AI agent for enhancement\nskill-seekers scrape --config configs\u002Freact.json --agent kimi\n```\n\n### PDF Extraction\n\n```bash\n# Basic PDF extraction\nskill-seekers pdf --pdf docs\u002Fmanual.pdf --name myskill\n\n# Advanced features\nskill-seekers pdf --pdf docs\u002Fmanual.pdf --name myskill \\\n    --extract-tables \\        # Extract tables\n    --parallel \\              # Fast parallel processing\n    --workers 8               # Use 8 CPU cores\n\n# Scanned PDFs (requires: pip install pytesseract Pillow)\nskill-seekers pdf --pdf docs\u002Fscanned.pdf --name myskill --ocr\n```\n\n### Video Extraction\n\n```bash\n# Install video support\npip install skill-seekers[video]        # Transcripts + metadata\npip install skill-seekers[video-full]   # + Whisper + visual frame extraction\n\n# Auto-detect GPU and install visual deps (PyTorch + easyocr)\nskill-seekers video --setup\n\n# Extract from YouTube video\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=dQw4w9WgXcQ --name mytutorial\n\n# Extract from a YouTube playlist\nskill-seekers video --playlist https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=... --name myplaylist\n\n# Extract from a local video file\nskill-seekers video --video-file recording.mp4 --name myrecording\n\n# Extract with visual frame analysis (requires video-full deps)\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --name mytutorial --visual\n\n# With AI enhancement (cleans OCR + generates polished SKILL.md)\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --visual --enhance-level 2\n\n# Clip a specific section of a video (supports seconds, MM:SS, HH:MM:SS)\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --start-time 1:30 --end-time 5:00\n\n# Use Vision API for low-confidence OCR frames (requires ANTHROPIC_API_KEY)\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --visual --vision-ocr\n\n# Re-build skill from previously extracted data (skip download)\nskill-seekers video --from-json output\u002Fmytutorial\u002Fvideo_data\u002Fextracted_data.json --name mytutorial\n```\n\n> **Full guide:** See [docs\u002FVIDEO_GUIDE.md](docs\u002FVIDEO_GUIDE.md) for complete CLI reference,\n> visual pipeline details, AI enhancement options, and troubleshooting.\n\n### GitHub Repository Analysis\n\n```bash\n# Basic repository scraping\nskill-seekers github --repo facebook\u002Freact\n\n# With authentication (higher rate limits)\nexport GITHUB_TOKEN=ghp_your_token_here\nskill-seekers github --repo facebook\u002Freact\n\n# Customize what to include\nskill-seekers github --repo django\u002Fdjango \\\n    --include-issues \\        # Extract GitHub Issues\n    --max-issues 100 \\        # Limit issue count\n    --include-changelog       # Extract CHANGELOG.md\n```\n\n### Unified Multi-Source Scraping\n\n**Combine documentation + GitHub + PDF into one unified skill with conflict detection:**\n\n```bash\n# Use existing unified configs\nskill-seekers unified --config configs\u002Freact_unified.json\nskill-seekers unified --config configs\u002Fdjango_unified.json\n\n# Or create unified config\ncat > configs\u002Fmyframework_unified.json \u003C\u003C 'EOF'\n{\n  \"name\": \"myframework\",\n  \"merge_mode\": \"rule-based\",\n  \"sources\": [\n    {\n      \"type\": \"documentation\",\n      \"base_url\": \"https:\u002F\u002Fdocs.myframework.com\u002F\",\n      \"max_pages\": 200\n    },\n    {\n      \"type\": \"github\",\n      \"repo\": \"owner\u002Fmyframework\",\n      \"code_analysis_depth\": \"surface\"\n    }\n  ]\n}\nEOF\n\nskill-seekers unified --config configs\u002Fmyframework_unified.json\n```\n\n**Conflict Detection automatically finds:**\n- 🔴 **Missing in code** (high): Documented but not implemented\n- 🟡 **Missing in docs** (medium): Implemented but not documented\n- ⚠️ **Signature mismatch**: Different parameters\u002Ftypes\n- ℹ️ **Description mismatch**: Different explanations\n\n**Full Guide:** See [docs\u002FUNIFIED_SCRAPING.md](docs\u002FUNIFIED_SCRAPING.md) for complete documentation.\n\n### Private Config Repositories\n\n**Share custom configs across teams using private git repositories:**\n\n```bash\n# Option 1: Using MCP tools (recommended)\n# Register your team's private repo\nadd_config_source(\n    name=\"team\",\n    git_url=\"https:\u002F\u002Fgithub.com\u002Fmycompany\u002Fskill-configs.git\",\n    token_env=\"GITHUB_TOKEN\"\n)\n\n# Fetch config from team repo\nfetch_config(source=\"team\", config_name=\"internal-api\")\n```\n\n**Supported Platforms:**\n- GitHub (`GITHUB_TOKEN`), GitLab (`GITLAB_TOKEN`), Gitea (`GITEA_TOKEN`), Bitbucket (`BITBUCKET_TOKEN`)\n\n**Full Guide:** See [docs\u002FGIT_CONFIG_SOURCES.md](docs\u002FGIT_CONFIG_SOURCES.md) for complete documentation.\n\n## How It Works\n\n```mermaid\ngraph LR\n    A[Documentation Website] --> B[Skill Seekers]\n    B --> C[Scraper]\n    B --> D[AI Enhancement]\n    B --> E[Packager]\n    C --> F[Organized References]\n    D --> F\n    F --> E\n    E --> G[AI Skill .zip]\n    G --> H[Upload to AI Platform]\n```\n\n0. **Detect llms.txt** - Checks for llms-full.txt, llms.txt, llms-small.txt first (part of Smart SPA Discovery)\n1. **Scrape**: Extracts all pages from documentation\n2. **Categorize**: Organizes content into topics (API, guides, tutorials, etc.)\n3. **Enhance**: AI analyzes docs and creates comprehensive SKILL.md with examples (supports multiple agents via `--agent`)\n4. **Package**: Bundles everything into a platform-ready `.zip` file\n\n## Architecture\n\nThe system is organized into **8 core modules** and **5 utility modules** (~200 classes total):\n\n![Package Overview](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_9101963a8489.png)\n\n| Module | Purpose | Key Classes |\n|--------|---------|-------------|\n| **CLICore** | Git-style command dispatcher | `CLIDispatcher`, `SourceDetector`, `CreateCommand` |\n| **Scrapers** | 17 source-type extractors | `DocToSkillConverter`, `GitHubScraper`, `UnifiedScraper` |\n| **Adaptors** | 20+ output platform formats | `SkillAdaptor` (ABC), `ClaudeAdaptor`, `LangChainAdaptor` |\n| **Analysis** | C3.x codebase analysis pipeline | `UnifiedCodebaseAnalyzer`, `PatternRecognizer`, 10 GoF detectors |\n| **Enhancement** | AI-powered skill improvement via `AgentClient` | `AgentClient`, `AIEnhancer`, `UnifiedEnhancer`, `WorkflowEngine` |\n| **Packaging** | Package, upload, install skills | `PackageSkill`, `InstallAgent` |\n| **MCP** | FastMCP server (40 tools) | `SkillSeekerMCPServer`, 10 tool modules |\n| **Sync** | Doc change detection | `ChangeDetector`, `SyncMonitor`, `Notifier` |\n\nUtility modules: **Parsers** (28 CLI parsers), **Storage** (S3\u002FGCS\u002FAzure), **Embedding** (multi-provider vectors), **Benchmark** (performance), **Utilities** (16 shared helpers).\n\nFull UML diagrams: **[docs\u002FUML_ARCHITECTURE.md](docs\u002FUML_ARCHITECTURE.md)** | StarUML project: `docs\u002FUML\u002Fskill_seekers.mdj` | HTML API reference: `docs\u002FUML\u002Fhtml\u002F`\n\n## 📋 Prerequisites\n\n**Before you start, make sure you have:**\n\n1. **Python 3.10 or higher** - [Download](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002F) | Check: `python3 --version`\n2. **Git** - [Download](https:\u002F\u002Fgit-scm.com\u002F) | Check: `git --version`\n3. **15-30 minutes** for first-time setup\n\n**First time user?** → **[Start Here: Bulletproof Quick Start Guide](BULLETPROOF_QUICKSTART.md)** 🎯\n\n---\n\n## 📤 Uploading Skills to Claude\n\nOnce your skill is packaged, you need to upload it to Claude:\n\n### Option 1: Automatic Upload (API-based)\n\n```bash\n# Set your API key (one-time)\nexport ANTHROPIC_API_KEY=sk-ant-...\n\n# Package and upload automatically\nskill-seekers package output\u002Freact\u002F --upload\n\n# OR upload existing .zip\nskill-seekers upload output\u002Freact.zip\n```\n\n### Option 2: Manual Upload (No API Key)\n\n```bash\n# Package skill\nskill-seekers package output\u002Freact\u002F\n# → Creates output\u002Freact.zip\n\n# Then manually upload:\n# - Go to https:\u002F\u002Fclaude.ai\u002Fskills\n# - Click \"Upload Skill\"\n# - Select output\u002Freact.zip\n```\n\n### Option 3: MCP (Claude Code)\n\n```\nIn Claude Code, just ask:\n\"Package and upload the React skill\"\n```\n\n---\n\n## 🤖 Installing to AI Agents\n\nSkill Seekers can automatically install skills to 18 AI coding agents.\n\n```bash\n# Install to specific agent\nskill-seekers install-agent output\u002Freact\u002F --agent cursor\n\n# Install to all agents at once\nskill-seekers install-agent output\u002Freact\u002F --agent all\n\n# Preview without installing\nskill-seekers install-agent output\u002Freact\u002F --agent cursor --dry-run\n```\n\n### Supported Agents\n\n| Agent | Path | Type |\n|-------|------|------|\n| **Claude Code** | `~\u002F.claude\u002Fskills\u002F` | Global |\n| **Cursor** | `.cursor\u002Fskills\u002F` | Project |\n| **VS Code \u002F Copilot** | `.github\u002Fskills\u002F` | Project |\n| **Amp** | `~\u002F.amp\u002Fskills\u002F` | Global |\n| **Goose** | `~\u002F.config\u002Fgoose\u002Fskills\u002F` | Global |\n| **OpenCode** | `~\u002F.opencode\u002Fskills\u002F` | Global |\n| **Windsurf** | `~\u002F.windsurf\u002Fskills\u002F` | Global |\n| **Roo Code** | `.roo\u002Fskills\u002F` | Project |\n| **Cline** | `.cline\u002Fskills\u002F` | Project |\n| **Aider** | `~\u002F.aider\u002Fskills\u002F` | Global |\n| **Bolt** | `.bolt\u002Fskills\u002F` | Project |\n| **Kilo Code** | `.kilo\u002Fskills\u002F` | Project |\n| **Continue** | `~\u002F.continue\u002Fskills\u002F` | Global |\n| **Kimi Code** | `~\u002F.kimi\u002Fskills\u002F` | Global |\n\n---\n\n## 🔌 MCP Integration (26 Tools)\n\nSkill Seekers ships an MCP server for use from Claude Code, Cursor, Windsurf, VS Code + Cline, or IntelliJ IDEA.\n\n```bash\n# stdio mode (Claude Code, VS Code + Cline)\npython -m skill_seekers.mcp.server_fastmcp\n\n# HTTP mode (Cursor, Windsurf, IntelliJ)\npython -m skill_seekers.mcp.server_fastmcp --transport http --port 8765\n\n# Auto-configure all agents at once\n.\u002Fsetup_mcp.sh\n```\n\n**All 26 tools available:**\n- **Core (9):** `list_configs`, `generate_config`, `validate_config`, `estimate_pages`, `scrape_docs`, `package_skill`, `upload_skill`, `enhance_skill`, `install_skill`\n- **Extended (10):** `scrape_github`, `scrape_pdf`, `unified_scrape`, `merge_sources`, `detect_conflicts`, `add_config_source`, `fetch_config`, `list_config_sources`, `remove_config_source`, `split_config`\n- **Vector DB (4):** `export_to_chroma`, `export_to_weaviate`, `export_to_faiss`, `export_to_qdrant`\n- **Cloud (3):** `cloud_upload`, `cloud_download`, `cloud_list`\n\n**Full Guide:** [docs\u002FMCP_SETUP.md](docs\u002FMCP_SETUP.md)\n\n---\n\n## ⚙️ Configuration\n\n### Available Presets (24+)\n\n```bash\n# List all presets\nskill-seekers list-configs\n```\n\n| Category | Presets |\n|----------|---------|\n| **Web Frameworks** | `react`, `vue`, `angular`, `svelte`, `nextjs` |\n| **Python** | `django`, `flask`, `fastapi`, `sqlalchemy`, `pytest` |\n| **Game Development** | `godot`, `pygame`, `unity` |\n| **Tools & DevOps** | `docker`, `kubernetes`, `terraform`, `ansible` |\n| **Unified (Docs + GitHub)** | `react-unified`, `vue-unified`, `nextjs-unified`, and more |\n\n### Creating Your Own Config\n\n```bash\n# Option 1: Interactive\nskill-seekers scrape --interactive\n\n# Option 2: Copy and edit a preset\ncp configs\u002Freact.json configs\u002Fmyframework.json\nnano configs\u002Fmyframework.json\nskill-seekers scrape --config configs\u002Fmyframework.json\n```\n\n### Config File Structure\n\n```json\n{\n  \"name\": \"myframework\",\n  \"description\": \"When to use this skill\",\n  \"base_url\": \"https:\u002F\u002Fdocs.myframework.com\u002F\",\n  \"selectors\": {\n    \"main_content\": \"article\",\n    \"title\": \"h1\",\n    \"code_blocks\": \"pre code\"\n  },\n  \"url_patterns\": {\n    \"include\": [\"\u002Fdocs\", \"\u002Fguide\"],\n    \"exclude\": [\"\u002Fblog\", \"\u002Fabout\"]\n  },\n  \"categories\": {\n    \"getting_started\": [\"intro\", \"quickstart\"],\n    \"api\": [\"api\", \"reference\"]\n  },\n  \"rate_limit\": 0.5,\n  \"max_pages\": 500\n}\n```\n\n### Where to Store Configs\n\nThe tool searches in this order:\n1. Exact path as provided\n2. `.\u002Fconfigs\u002F` (current directory)\n3. `~\u002F.config\u002Fskill-seekers\u002Fconfigs\u002F` (user config directory)\n4. SkillSeekersWeb.com API (preset configs)\n\n---\n\n## 📊 What Gets Created\n\n```\noutput\u002F\n├── godot_data\u002F              # Scraped raw data\n│   ├── pages\u002F              # JSON files (one per page)\n│   └── summary.json        # Overview\n│\n└── godot\u002F                   # The skill\n    ├── SKILL.md            # Enhanced with real examples\n    ├── references\u002F         # Categorized docs\n    │   ├── index.md\n    │   ├── getting_started.md\n    │   ├── scripting.md\n    │   └── ...\n    ├── scripts\u002F            # Empty (add your own)\n    └── assets\u002F             # Empty (add your own)\n```\n\n---\n\n## 🐛 Troubleshooting\n\n### No Content Extracted?\n- Check your `main_content` selector\n- Try: `article`, `main`, `div[role=\"main\"]`\n\n### Data Exists But Won't Use It?\n```bash\n# Force re-scrape\nrm -rf output\u002Fmyframework_data\u002F\nskill-seekers scrape --config configs\u002Fmyframework.json\n```\n\n### Categories Not Good?\nEdit the config `categories` section with better keywords.\n\n### Want to Update Docs?\n```bash\n# Delete old data and re-scrape\nrm -rf output\u002Fgodot_data\u002F\nskill-seekers scrape --config configs\u002Fgodot.json\n```\n\n### Enhancement Not Working?\n```bash\n# Check if API key is set\necho $ANTHROPIC_API_KEY\n\n# Try LOCAL mode instead (uses Claude Code Max, no API key needed)\nskill-seekers enhance output\u002Freact\u002F --mode LOCAL\n\n# Monitor background enhancement status\nskill-seekers enhance-status output\u002Freact\u002F --watch\n```\n\n### GitHub Rate Limit Issues?\n```bash\n# Set a GitHub token (5000 req\u002Fhour vs 60\u002Fhour anonymous)\nexport GITHUB_TOKEN=ghp_your_token_here\n\n# Or configure multiple profiles\nskill-seekers config --github\n```\n\n---\n\n## 📈 Performance\n\n| Task | Time | Notes |\n|------|------|-------|\n| Scraping (sync) | 15-45 min | First time only, thread-based |\n| Scraping (async) | 5-15 min | 2-3x faster with `--async` flag |\n| Building | 1-3 min | Fast rebuild from cache |\n| Re-building | \u003C1 min | With `--skip-scrape` |\n| Enhancement (LOCAL) | 30-60 sec | Uses Claude Code Max |\n| Enhancement (API) | 20-40 sec | Requires API key |\n| Video (transcript) | 1-3 min | YouTube\u002Flocal, transcript only |\n| Video (visual) | 5-15 min | + OCR frame extraction |\n| Packaging | 5-10 sec | Final .zip creation |\n\n---\n\n## 📚 Documentation\n\n### Getting Started\n- **[BULLETPROOF_QUICKSTART.md](BULLETPROOF_QUICKSTART.md)** - 🎯 **START HERE** if you're new!\n- **[QUICKSTART.md](QUICKSTART.md)** - Quick start for experienced users\n- **[TROUBLESHOOTING.md](TROUBLESHOOTING.md)** - Common issues and solutions\n- **[docs\u002FQUICK_REFERENCE.md](docs\u002FQUICK_REFERENCE.md)** - One-page cheat sheet\n\n### Architecture\n- **[docs\u002FUML_ARCHITECTURE.md](docs\u002FUML_ARCHITECTURE.md)** - UML architecture overview with 14 diagrams\n- **[docs\u002FUML\u002Fexports\u002F](docs\u002FUML\u002Fexports\u002F)** - PNG diagram exports (package overview + 13 class diagrams)\n- **[docs\u002FUML\u002Fhtml\u002F](docs\u002FUML\u002Fhtml\u002Findex.html\u002Findex.html)** - Full HTML API reference (all classes, operations, attributes)\n- **[docs\u002FUML\u002Fskill_seekers.mdj](docs\u002FUML\u002Fskill_seekers.mdj)** - StarUML project file (open with [StarUML](https:\u002F\u002Fstaruml.io\u002F))\n\n### Guides\n- **[docs\u002FLARGE_DOCUMENTATION.md](docs\u002FLARGE_DOCUMENTATION.md)** - Handle 10K-40K+ page docs\n- **[ASYNC_SUPPORT.md](ASYNC_SUPPORT.md)** - Async mode guide (2-3x faster scraping)\n- **[docs\u002FENHANCEMENT_MODES.md](docs\u002FENHANCEMENT_MODES.md)** - AI enhancement modes guide\n- **[docs\u002FMCP_SETUP.md](docs\u002FMCP_SETUP.md)** - MCP integration setup\n- **[docs\u002FUNIFIED_SCRAPING.md](docs\u002FUNIFIED_SCRAPING.md)** - Multi-source scraping\n- **[docs\u002FVIDEO_GUIDE.md](docs\u002FVIDEO_GUIDE.md)** - Video extraction guide\n\n### Integration Guides\n- **[docs\u002Fintegrations\u002FLANGCHAIN.md](docs\u002Fintegrations\u002FLANGCHAIN.md)** - LangChain RAG\n- **[docs\u002Fintegrations\u002FCURSOR.md](docs\u002Fintegrations\u002FCURSOR.md)** - Cursor IDE\n- **[docs\u002Fintegrations\u002FWINDSURF.md](docs\u002Fintegrations\u002FWINDSURF.md)** - Windsurf IDE\n- **[docs\u002Fintegrations\u002FCLINE.md](docs\u002Fintegrations\u002FCLINE.md)** - Cline (VS Code)\n- **[docs\u002Fintegrations\u002FRAG_PIPELINES.md](docs\u002Fintegrations\u002FRAG_PIPELINES.md)** - All RAG pipelines\n\n---\n\n## 📝 License\n\nMIT License - see [LICENSE](LICENSE) file for details\n\n---\n\nHappy skill building! 🚀\n\n---\n\n## 🔒 Security\n\n[![MseeP.ai Security Assessment Badge](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_f0c1c269426f.png)](https:\u002F\u002Fmseep.ai\u002Fapp\u002Fyusufkaraaslan-skill-seekers)\n","\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_86f7712e0c1b.png\" alt=\"技能探索者\" width=\"200\"\u002F>\n\u003C\u002Fp>\n\n# 技能探索者\n\nEnglish | [简体中文](README.zh-CN.md) | [日本語](README.ja.md) | [한국어](README.ko.md) | [Español](README.es.md) | [Français](README.fr.md) | [Deutsch](README.de.md) | [Português](README.pt-BR.md) | [Türkçe](README.tr.md) | [العربية](README.ar.md) | [हिन्दी](README.hi.md) | [Русский](README.ru.md)\n\n[![版本](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fversion-3.5.0-blue.svg)](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Freleases)\n[![许可证: MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-yellow.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n[![Python 3.10+](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.10+-blue.svg)](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002F)\n[![MCP集成](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMCP-Integrated-blue.svg)](https:\u002F\u002Fmodelcontextprotocol.io)\n[![已测试](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTests-2540%2B%20Passing-brightgreen.svg)](tests\u002F)\n[![项目看板](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FProject-Board-purple.svg)](https:\u002F\u002Fgithub.com\u002Fusers\u002Fyusufkaraaslan\u002Fprojects\u002F2)\n[![PyPI版本](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Fskill-seekers.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F)\n[![PyPI - 下载量](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fskill-seekers.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F)\n[![PyPI - Python版本](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fskill-seekers.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F)\n[![官网](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWebsite-skillseekersweb.com-blue.svg)](https:\u002F\u002Fskillseekersweb.com\u002F)\n[![Twitter关注](https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Ffollow\u002F_yUSyUS_?style=social)](https:\u002F\u002Fx.com\u002F_yUSyUS_)\n[![GitHub仓库星标数](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fyusufkaraaslan\u002FSkill_Seekers?style=social)](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers)\n[![PyPI下载量](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_0cc707d5ff49.png)](https:\u002F\u002Fpepy.tech\u002Fprojects\u002Fskill-seekers)\n\n\u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F18329\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_1145cd82417e.png\" alt=\"yusufkaraaslan%2FSkill_Seekers | Trendshift\" style=\"width: 250px; height: 55px;\" width=\"250\" height=\"55\"\u002F>\u003C\u002Fa>\n\n**🧠 人工智能系统的数据层。** Skill Seekers 可以将文档网站、GitHub 仓库、PDF 文件、视频、笔记本、维基以及其他 10 多种来源类型转化为结构化的知识资产，从而在几分钟内而非几小时内为 AI 技能（Claude、Gemini、OpenAI）、RAG 流水线（LangChain、LlamaIndex、Pinecone）以及 AI 编码助手（Cursor、Windsurf、Cline）提供支持。\n\n> 🌐 **[访问 SkillSeekersWeb.com](https:\u002F\u002Fskillseekersweb.com\u002F)** - 浏览 24 种以上的预设配置，分享您的配置，并获取完整的文档！\n\n> 📋 **[查看开发路线图与任务](https:\u002F\u002Fgithub.com\u002Fusers\u002Fyusufkaraaslan\u002Fprojects\u002F2)** - 10 个类别中共有 134 项任务，您可以选择任意一项参与贡献！\n\n## 🌐 生态系统\n\nSkill Seekers 是一个多仓库项目。以下是各个项目的存放位置：\n\n| 仓库 | 描述 | 链接 |\n|-----------|-------------|-------|\n| **[Skill_Seekers](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers)** | 核心 CLI 和 MCP 服务器（本仓库） | [PyPI](https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F) |\n| **[skillseekersweb](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskillseekersweb)** | 网站和文档 | [在线](https:\u002F\u002Fskillseekersweb.com\u002F) |\n| **[skill-seekers-configs](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskill-seekers-configs)** | 社区配置仓库 | |\n| **[skill-seekers-action](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskill-seekers-action)** | 用于 CI\u002FCD 的 GitHub Action | |\n| **[skill-seekers-plugin](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskill-seekers-plugin)** | Claude Code 插件 | |\n| **[homebrew-skill-seekers](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fhomebrew-skill-seekers)** | macOS 的 Homebrew tap | |\n\n> **想参与贡献吗？** 网站和配置仓库是新贡献者的好起点！\n\n## 🧠 人工智能系统的数据层\n\n**Skill Seekers 是通用的预处理层**，位于原始文档和所有消费这些文档的人工智能系统之间。无论您是在构建 Claude 技能、LangChain RAG 流水线，还是一份 Cursor `.cursorrules` 文件——数据准备过程都是相同的。您只需进行一次准备，即可导出到所有目标平台。\n\n```bash\n# 一条命令 → 结构化知识资产\nskill-seekers create https:\u002F\u002Fdocs.react.dev\u002F\n# 或：skill-seekers create facebook\u002Freact\n# 或：skill-seekers create .\u002Fmy-project\n\n# 导出到任何人工智能系统\nskill-seekers package output\u002Freact --target claude      # → Claude AI 技能 (ZIP)\nskill-seekers package output\u002Freact --target langchain   # → LangChain 文档\nskill-seekers package output\u002Freact --target llama-index # → LlamaIndex TextNodes\nskill-seekers package output\u002Freact --target cursor      # → .cursorrules\n```\n\n### 构建的内容\n\n| 输出 | 目标 | 支持的内容 |\n|--------|--------|---------------|\n| **Claude 技能**（ZIP + YAML） | `--target claude` | Claude Code, Claude API |\n| **Gemini 技能**（tar.gz） | `--target gemini` | Google Gemini |\n| **OpenAI \u002F 自定义 GPT**（ZIP） | `--target openai` | GPT-4o、自定义助手 |\n| **LangChain 文档** | `--target langchain` | QA 链、代理、检索器 |\n| **LlamaIndex TextNodes** | `--target llama-index` | 查询引擎、聊天引擎 |\n| **Haystack 文档** | `--target haystack` | 企业级 RAG 流水线 |\n| **Pinecone 就绪**（Markdown） | `--target markdown` | 向量插入 |\n| **ChromaDB \u002F FAISS \u002F Qdrant** | `--format chroma\u002Ffaiss\u002Fqdrant` | 本地向量数据库 |\n| **Cursor** `.cursorrules` | `--target claude` → 复制 | Cursor IDE 的 AI 上下文 |\n| **Windsurf \u002F Cline \u002F Continue** | `--target claude` → 复制 | VS Code、IntelliJ、Vim |\n\n### 为什么重要\n\n- ⚡ **快 99%** — 数天的手动数据准备 → 15–45 分钟\n- 🎯 **AI 技能质量** — 500 多行的 SKILL.md 文件，包含示例、模式和指南\n- 📊 **适合 RAG 的分块** — 智能分块保留代码块并维持上下文\n- 🎬 **视频** — 从 YouTube 和本地视频中提取代码、字幕和结构化的知识\n- 🔄 **多源整合** — 将 17 种来源类型（文档、GitHub、PDF、视频、笔记本、维基等）合并为一个知识资产\n- 🌐 **一次准备，多平台适用** — 将同一资产导出到 16 个平台，无需重新抓取\n- ✅ **经过实战检验** — 2,540 多项测试、24 多种框架预设，可直接投入生产\n\n## 🚀 快速入门（3 条命令）\n\n```bash\n# 1. 安装\npip install skill-seekers\n\n# 2. 从任何来源创建技能\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F\n\n# 3. 打包以供您的 AI 平台使用\nskill-seekers package output\u002Fdjango --target claude\n```\n\n**就是这样！** 您现在拥有了 `output\u002Fdjango-claude.zip`，可以立即使用。\n\n```bash\n# 使用不同的 AI 代理进行增强（默认：claude）\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F --agent kimi\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F --agent codex\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F --agent-cmd \"my-custom-agent run\"\n```\n\n### 其他来源（支持 17 种）\n\n```bash\n\n# GitHub 仓库\nskill-seekers create facebook\u002Freact\n\n# 本地项目\nskill-seekers create .\u002Fmy-project\n\n# PDF 文档\nskill-seekers create manual.pdf\n\n# Word 文档\nskill-seekers create report.docx\n\n# EPUB 电子书\nskill-seekers create book.epub\n\n# Jupyter Notebook\nskill-seekers create notebook.ipynb\n\n# OpenAPI 规范\nskill-seekers create openapi.yaml\n\n# PowerPoint 演示文稿\nskill-seekers create presentation.pptx\n\n# AsciiDoc 文档\nskill-seekers create guide.adoc\n\n# 本地 HTML 文件\nskill-seekers create page.html\n\n# RSS\u002FAtom 订阅源\nskill-seekers create feed.rss\n\n# 手册页\nskill-seekers create curl.1\n\n# 视频（YouTube、Vimeo 或本地文件 — 需要 skill-seekers[video]）\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --name mytutorial\n# 第一次使用？自动安装支持 GPU 的可视化依赖：\nskill-seekers video --setup\n\n# Confluence 维基\nskill-seekers confluence --space TEAM --name wiki\n\n# Notion 页面\nskill-seekers notion --database-id ... --name docs\n\n# Slack\u002FDiscord 聊天记录导出\nskill-seekers chat --export-dir .\u002Fslack-export --name team-chat\n```\n\n### 导出到各平台\n\n```bash\n# 针对多个平台的打包\nfor platform in claude gemini openai langchain; do\n  skill-seekers package output\u002Fdjango --target $platform\ndone\n```\n\n## Skill Seekers 是什么？\n\nSkill Seekers 是 **AI 系统的数据层**。它能够将 17 种来源类型——文档网站、GitHub 仓库、PDF、视频、Jupyter Notebook、Word\u002FEPUB\u002FAsciiDoc 文档、OpenAPI 规范、PowerPoint 演示文稿、RSS 订阅源、手册页、Confluence 维基、Notion 页面、Slack\u002FDiscord 导出等——转化为结构化的知识资产，供各类 AI 目标使用：\n\n| 使用场景 | 输出内容 | 示例 |\n|----------|-------------|---------|\n| **AI 技能** | 全面的 SKILL.md + 引用 | Claude Code、Gemini、GPT |\n| **RAG 流水线** | 带有丰富元数据的分块文档 | LangChain、LlamaIndex、Haystack |\n| **向量数据库** | 已格式化并可直接插入的数据 | Pinecone、Chroma、Weaviate、FAISS |\n| **AI 编程助手** | IDE 中的 AI 可自动读取的上下文文件 | Cursor、Windsurf、Cline、Continue.dev |\n\n## 📚 文档\n\n| 我想... | 阅读这篇 |\n|--------------|-----------|\n| **快速入门** | [快速入门](docs\u002Fgetting-started\u002F02-quick-start.md) - 3 条命令即可获得首个技能 |\n| **理解概念** | [核心概念](docs\u002Fuser-guide\u002F01-core-concepts.md) - 工作原理 |\n| **抓取源数据** | [抓取指南](docs\u002Fuser-guide\u002F02-scraping.md) - 所有源类型 |\n| **增强技能** | [增强指南](docs\u002Fuser-guide\u002F03-enhancement.md) - AI 增强 |\n| **导出技能** | [打包指南](docs\u002Fuser-guide\u002F04-packaging.md) - 平台导出 |\n| **查找命令** | [CLI 参考](docs\u002Freference\u002FCLI_REFERENCE.md) - 所有 20 条命令 |\n| **配置** | [配置格式](docs\u002Freference\u002FCONFIG_FORMAT.md) - JSON 规范 |\n| **故障排除** | [故障排除](docs\u002Fuser-guide\u002F06-troubleshooting.md) - 常见问题 |\n\n**完整文档：** [docs\u002FREADME.md](docs\u002FREADME.md)\n\n与其花费数天进行手动预处理，Skill Seekers 能够：\n\n1. **摄取** — 文档、GitHub 仓库、本地代码库、PDF、视频、Notebook、维基等 10 多种来源\n2. **分析** — 深度 AST 解析、模式检测、API 提取\n3. **结构化** — 分类参考文件与元数据\n4. **增强** — AI 驱动的 SKILL.md 生成（Claude、Gemini 或本地模型）\n5. **导出** — 从一份资产中导出 16 种平台特定的格式\n\n## 为什么使用它？\n\n### 对于 AI 技能构建者（Claude、Gemini、OpenAI）\n\n- 🎯 **生产级技能** — 500 行以上的 SKILL.md 文件，包含代码示例、模式和指南\n- 🔄 **增强工作流** — 应用 `security-focus`、`architecture-comprehensive` 或自定义 YAML 预设\n- 🎮 **任意领域** — 游戏引擎（Godot、Unity）、框架（React、Django）、内部工具\n- 🔧 **团队** — 将内部文档与代码整合为单一的事实来源\n- 📚 **高质量** — AI 增强，附带示例、快速参考和导航指引\n\n### 对于 RAG 构建者及 AI 工程师\n\n- 🤖 **适合 RAG 的数据** — 预先分块的 LangChain `Documents`、LlamaIndex `TextNodes`、Haystack `Documents`\n- 🚀 **快 99%** — 数天的预处理 → 15–45 分钟\n- 📊 **智能元数据** — 分类、来源、类型 → 更高的检索准确率\n- 🔄 **多源整合** — 将文档、GitHub、PDF 和视频整合到一条流水线中\n- 🌐 **平台无关** — 可导出至任何向量数据库或框架，无需重新抓取\n\n### 对于 AI 编程助手用户\n\n- 💻 **Cursor \u002F Windsurf \u002F Cline** — 自动生成 `.cursorrules` \u002F `.windsurfrules` \u002F `.clinerules`\n- 🎯 **持久上下文** — AI“了解”你的框架，无需反复提示\n- 📚 **始终最新** — 当文档更新时，可在几分钟内更新上下文\n\n## 核心功能\n\n### 🌐 文档抓取\n- ✅ **智能 SPA 发现** - 三层发现机制，适用于 JavaScript SPA 网站（sitemap.xml → llms.txt → 无头浏览器渲染）\n- ✅ **llms.txt 支持** - 自动检测并使用 LLM 就绪的文档文件（速度提升 10 倍）\n- ✅ **通用抓取器** - 适用于任何文档网站\n- ✅ **智能分类** - 自动按主题组织内容\n- ✅ **代码语言检测** - 识别 Python、JavaScript、C++、GDScript 等\n- ✅ **24+ 即用预设** - Godot、React、Vue、Django、FastAPI 等\n\n### 📄 PDF 支持\n- ✅ **基础 PDF 提取** - 从 PDF 文件中提取文本、代码和图片\n- ✅ **扫描 PDF 的 OCR** - 从扫描文档中提取文本\n- ✅ **密码保护的 PDF** - 处理加密 PDF\n- ✅ **表格提取** - 从 PDF 中提取复杂表格\n- ✅ **并行处理** - 大型 PDF 处理速度提升 3 倍\n- ✅ **智能缓存** - 再次运行时速度提升 50%\n\n### 🎬 视频提取\n- ✅ **YouTube 和本地视频** - 从视频中提取字幕、屏幕上的代码以及结构化知识\n- ✅ **视觉帧分析** - 从代码编辑器、终端、幻灯片和图表中提取 OCR 数据\n- ✅ **GPU 自动检测** - 自动安装正确的 PyTorch 版本（CUDA\u002FROCm\u002FMPS\u002FCPU）\n- ✅ **AI 增强** - 两步流程：清除 OCR 杂质 + 生成精炼的 SKILL.md\n- ✅ **时间剪辑** - 使用 `--start-time` 和 `--end-time` 提取特定片段\n- ✅ **播放列表支持** - 批量处理 YouTube 播放列表中的所有视频\n- ✅ **Vision API 备用** - 在 OCR 置信度较低时使用 Claude Vision\n\n### 🐙 GitHub 仓库分析\n- ✅ **深度代码分析** - 对 Python、JavaScript、TypeScript、Java、C++、Go 进行 AST 解析\n- ✅ **API 提取** - 函数、类、方法及其参数和类型\n- ✅ **仓库元数据** - README、文件树、语言分布、星标\u002F分支数\n- ✅ **GitHub 问题和 PR** - 获取已关闭和未解决的问题，附带标签和里程碑\n- ✅ **CHANGELOG 和发布** - 自动提取版本历史\n- ✅ **冲突检测** - 比较文档中的 API 与实际代码实现\n- ✅ **MCP 集成** - 自然语言输入：“抓取 GitHub 仓库 facebook\u002Freact”\n\n### 🔄 统一多源抓取\n- ✅ **整合多个来源** - 将文档、GitHub 和 PDF 混合到一个技能中\n- ✅ **冲突检测** - 自动发现文档与代码之间的不一致\n- ✅ **智能合并** - 基于规则或 AI 驱动的冲突解决\n- ✅ **透明报告** - 并排对比，并附带 ⚠️ 警告\n- ✅ **文档缺口分析** - 识别过时的文档和未记录的功能\n- ✅ **单一事实来源** - 一个技能同时展示意图（文档）和现实（代码）\n- ✅ **向后兼容** - 旧版单源配置仍可正常工作\n\n### 🤖 多 LLM 平台支持\n- ✅ **12 个 LLM 平台** - Claude AI、Google Gemini、OpenAI ChatGPT、MiniMax AI、通用 Markdown、OpenCode、Kimi（Moonshot AI）、DeepSeek AI、Qwen（阿里巴巴）、OpenRouter、Together AI、Fireworks AI\n- ✅ **通用抓取** - 同一份文档适用于所有平台\n- ✅ **平台特定打包** - 针对每个 LLM 优化格式\n- ✅ **一键导出** - 使用 `--target` 标志选择平台\n- ✅ **可选依赖** - 只安装你需要的部分\n- ✅ **100% 向后兼容** - 现有 Claude 工作流无需更改\n\n| 平台 | 格式 | 上传 | 增强 | API 密钥 | 自定义端点 |\n|----------|--------|--------|-------------|---------|-----------------|\n| **Claude AI** | ZIP + YAML | ✅ 自动 | ✅ 是 | ANTHROPIC_API_KEY | ANTHROPIC_BASE_URL |\n| **Google Gemini** | tar.gz | ✅ 自动 | ✅ 是 | GOOGLE_API_KEY | - |\n| **OpenAI ChatGPT** | ZIP + 向量存储 | ✅ 自动 | ✅ 是 | OPENAI_API_KEY | - |\n| **MiniMax AI** | ZIP + 知识文件 | ✅ 自动 | ✅ 是 | MINIMAX_API_KEY | - |\n| **通用 Markdown** | ZIP | ❌ 手动 | ❌ 否 | - | - |\n\n```bash\n# Claude（默认，无需更改！）\nskill-seekers package output\u002Freact\u002F\nskill-seekers upload react.zip\n\n# Google Gemini\npip install skill-seekers[gemini]\nskill-seekers package output\u002Freact\u002F --target gemini\nskill-seekers upload react-gemini.tar.gz --target gemini\n\n# OpenAI ChatGPT\npip install skill-seekers[openai]\nskill-seekers package output\u002Freact\u002F --target openai\nskill-seekers upload react-openai.zip --target openai\n\n# MiniMax AI\npip install skill-seekers[minimax]\nskill-seekers package output\u002Freact\u002F --target minimax\nskill-seekers upload react-minimax.zip --target minimax\n\n# 通用 Markdown（通用导出）\nskill-seekers package output\u002Freact\u002F --target markdown\n# 直接在任何 LLM 中使用这些 Markdown 文件\n```\n\n\u003Cdetails>\n\u003Csummary>🔧 \u003Cstrong>Claude 兼容 API 的环境变量（例如 GLM-4.7）\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nSkill Seekers 支持任何 Claude 兼容的 API 端点：\n\n```bash\n# 选项 1：官方 Anthropic API（默认）\nexport ANTHROPIC_API_KEY=sk-ant-...\n\n# 选项 2：GLM-4.7 Claude 兼容 API\nexport ANTHROPIC_API_KEY=your-glm-47-api-key\nexport ANTHROPIC_BASE_URL=https:\u002F\u002Fglm-4-7-endpoint.com\u002Fv1\n\n# 所有 AI 增强功能都将使用配置的端点\nskill-seekers enhance output\u002Freact\u002F\nskill-seekers analyze --directory . --enhance\n```\n\n**注意**：设置 `ANTHROPIC_BASE_URL` 允许你使用任何 Claude 兼容的 API 端点，比如 GLM-4.7（智谱 AI）或其他兼容服务。\n\n\u003C\u002Fdetails>\n\n**安装：**\n```bash\n# 安装支持 Gemini 的版本\npip install skill-seekers[gemini]\n\n# 安装支持 OpenAI 的版本\npip install skill-seekers[openai]\n\n# 安装支持 MiniMax 的版本\npip install skill-seekers[minimax]\n\n# 安装支持所有 LLM 平台的版本\npip install skill-seekers[all-llms]\n```\n\n### 🔗 RAG 框架集成\n\n- ✅ **LangChain Documents** - 直接导出为 `Document` 格式，包含 `page_content` 和元数据\n  - 非常适合：问答链、检索器、向量存储、代理\n  - 示例：[LangChain RAG 流程](examples\u002Flangchain-rag-pipeline\u002F)\n  - 指南：[LangChain 集成](docs\u002Fintegrations\u002FLANGCHAIN.md)\n\n- ✅ **LlamaIndex TextNodes** - 导出为 `TextNode` 格式，带有唯一 ID 和嵌入\n  - 非常适合：查询引擎、聊天引擎、存储上下文\n  - 示例：[LlamaIndex 查询引擎](examples\u002Fllama-index-query-engine\u002F)\n  - 指南：[LlamaIndex 集成](docs\u002Fintegrations\u002FLLAMA_INDEX.md)\n\n- ✅ **Pinecone 就绪格式** - 优化用于向量数据库的插入更新操作\n  - 非常适合：生产级向量搜索、语义搜索、混合搜索\n  - 示例：[Pinecone 插入更新](examples\u002Fpinecone-upsert\u002F)\n  - 指南：[Pinecone 集成](docs\u002Fintegrations\u002FPINECONE.md)\n\n**快速导出：**\n```bash\n# LangChain Documents（JSON）\nskill-seekers package output\u002Fdjango --target langchain\n# → output\u002Fdjango-langchain.json\n\n# LlamaIndex TextNodes（JSON）\nskill-seekers package output\u002Fdjango --target llama-index\n# → output\u002Fdjango-llama-index.json\n\n# Markdown（通用）\nskill-seekers package output\u002Fdjango --target markdown\n# → output\u002Fdjango-markdown\u002FSKILL.md + references\u002F\n```\n\n**完整的 RAG 流程指南**：[RAG 流程文档](docs\u002Fintegrations\u002FRAG_PIPELINES.md)\n\n---\n\n### 🧠 AI 编码助手集成\n\n将任何框架的文档转化为专家级编码上下文，供 4 种以上的 AI 助手使用：\n\n- ✅ **Cursor IDE** - 生成 `.cursorrules` 文件，用于 AI 驱动的代码建议\n  - 非常适合：框架特定的代码生成、一致性模式\n  - 适用：Cursor IDE（VS Code 分支）\n  - 指南：[Cursor 集成](docs\u002Fintegrations\u002FCURSOR.md)\n  - 示例：[Cursor React 技能](examples\u002Fcursor-react-skill\u002F)\n\n- ✅ **Windsurf** - 使用 `.windsurfrules` 自定义 Windsurf 的 AI 助手上下文\n  - 非常适合：IDE 原生 AI 辅助、基于流程的编码\n  - 适用：Codeium 的 Windsurf IDE\n  - 指南：[Windsurf 集成](docs\u002Fintegrations\u002FWINDSURF.md)\n  - 示例：[Windsurf FastAPI 上下文](examples\u002Fwindsurf-fastapi-context\u002F)\n\n- ✅ **Cline（VS Code）** - 系统提示 + MCP，用于 VS Code 代理\n  - 非常适合：在 VS Code 中进行代理式代码生成\n  - 适用：Cline 扩展（VS Code）\n  - 指南：[Cline 集成](docs\u002Fintegrations\u002FCLINE.md)\n  - 示例：[Cline Django 助手](examples\u002Fcline-django-assistant\u002F)\n\n- ✅ **Continue.dev** - 为跨 IDE 的 AI 提供上下文服务器\n  - 非常适合：多 IDE 环境（VS Code、JetBrains、Vim），自定义 LLM 提供者\n  - 适用：任何安装了 Continue.dev 插件的 IDE\n  - 指南：[Continue 集成](docs\u002Fintegrations\u002FCONTINUE_DEV.md)\n  - 示例：[Continue 通用上下文](examples\u002Fcontinue-dev-universal\u002F)\n\n**针对 AI 编码工具的快速导出：**\n```bash\n# 适用于任何 AI 编码助手（Cursor、Windsurf、Cline、Continue.dev）\nskill-seekers scrape --config configs\u002Fdjango.json\nskill-seekers package output\u002Fdjango --target claude  # 或 --target markdown\n\n# 复制到你的项目中（以 Cursor 为例）\ncp output\u002Fdjango-claude\u002FSKILL.md my-project\u002F.cursorrules\n\n# 或者用于 Windsurf\ncp output\u002Fdjango-claude\u002FSKILL.md my-project\u002F.windsurf\u002Frules\u002Fdjango.md\n\n# 或者用于 Cline\ncp output\u002Fdjango-claude\u002FSKILL.md my-project\u002F.clinerules\n\n# 或者用于 Continue.dev（HTTP 服务器）\npython examples\u002Fcontinue-dev-universal\u002Fcontext_server.py\n# 在 ~\u002F.continue\u002Fconfig.json 中进行配置\n```\n\n**集成中心**：[所有 AI 系统集成](docs\u002Fintegrations\u002FINTEGRATIONS.md)\n\n---\n\n### 🌊 三流 GitHub 架构\n- ✅ **三重流分析** - 将 GitHub 仓库拆分为代码、文档和洞察三个流\n- ✅ **统一代码库分析器** - 支持 GitHub URL 和本地路径\n- ✅ **C3.x 作为分析深度** - 可选择“basic”（1-2 分钟）或“c3x”（20-60 分钟）分析\n- ✅ **增强的路由生成** - 包含 GitHub 元数据、README 快速入门指南及常见问题\n- ✅ **问题集成** - 从 GitHub 问题中提取顶级问题及其解决方案\n- ✅ **智能路由关键词** - GitHub 标签权重提升 2 倍，以更好地检测主题\n\n**三流详解：**\n- **流 1：代码** - 深度 C3.x 分析（模式、示例、指南、配置、架构）\n- **流 2：文档** - 仓库文档（README、CONTRIBUTING、docs\u002F*.md）\n- **流 3：洞察** - 社区知识（问题、标签、星标、分支）\n\n```python\nfrom skill_seekers.cli.unified_codebase_analyzer import UnifiedCodebaseAnalyzer\n\n# 使用所有三流分析 GitHub 仓库\nanalyzer = UnifiedCodebaseAnalyzer()\nresult = analyzer.analyze(\n    source=\"https:\u002F\u002Fgithub.com\u002Ffacebook\u002Freact\",\n    depth=\"c3x\",  # 或 \"basic\" 进行快速分析\n    fetch_github_metadata=True\n)\n\n# 访问代码流（C3.x 分析）\nprint(f\"设计模式: {len(result.code_analysis['c3_1_patterns'])}\")\nprint(f\"测试示例: {result.code_analysis['c3_2_examples_count']}\")\n\n# 访问文档流（仓库文档）\nprint(f\"README: {result.github_docs['readme'][:100]}\")\n\n# 访问洞察流（GitHub 元数据）\nprint(f\"星标数: {result.github_insights['metadata']['stars']}\")\nprint(f\"常见问题: {len(result.github_insights['common_problems'])}\")\n```\n\n**查看完整文档**：[三流实现摘要](docs\u002FIMPLEMENTATION_SUMMARY_THREE_STREAM.md)\n\n### 🔐 智能速率限制管理与配置\n- ✅ **多令牌配置系统** - 管理多个 GitHub 账户（个人、工作、开源项目）\n  - 安全配置存储于 `~\u002F.config\u002Fskill-seekers\u002Fconfig.json`（权限 600）\n  - 每个账户的速率限制策略：提示、等待、切换、失败\n  - 可配置每个账户的超时时间（默认 30 分钟，防止无限等待）\n  - 智能回退链：CLI 参数 → 环境变量 → 配置文件 → 提示\n  - 支持 Claude、Gemini、OpenAI 的 API 密钥管理\n- ✅ **交互式配置向导** - 美观的终端 UI，便于设置\n  - 浏览器集成用于创建令牌（自动打开 GitHub 等）\n  - 令牌验证和连接测试\n  - 带颜色编码的可视化状态显示\n- ✅ **智能速率限制处理器** - 再也不用无限等待了！\n  - 提前警告速率限制情况（60\u002F小时 vs 5000\u002F小时）\n  - 实时从 GitHub API 响应中检测\n  - 带进度的倒计时定时器\n  - 当达到速率限制时自动切换账户\n  - 四种策略：提示（询问）、等待（倒计时）、切换（尝试其他账户）、失败（终止）\n- ✅ **恢复功能** - 继续中断的任务\n  - 按可配置间隔自动保存进度（默认 60 秒）\n  - 列出所有可恢复任务及其进度详情\n  - 自动清理旧任务（默认 7 天）\n- ✅ **CI\u002FCD 支持** - 非交互模式，适用于自动化\n  - `--non-interactive` 标志会快速失败，不弹出提示\n  - `--profile` 标志用于选择特定 GitHub 账户\n  - 清晰的错误信息，便于流水线日志记录\n\n**快速设置：**\n```bash\n# 一次性配置（5 分钟）\nskill-seekers config --github\n\n# 使用特定账户访问私有仓库\nskill-seekers github --repo mycompany\u002Fprivate-repo --profile work\n\n# CI\u002FCD 模式（快速失败，无提示）\nskill-seekers github --repo owner\u002Frepo --non-interactive\n\n# 恢复中断的任务\nskill-seekers resume --list\nskill-seekers resume github_react_20260117_143022\n```\n\n**速率限制策略说明：**\n- **prompt**（默认） - 当达到速率限制时询问如何处理（等待、切换、设置令牌、取消）\n- **wait** - 自动等待并显示倒计时（尊重超时设置）\n- **switch** - 自动尝试下一个可用账户（适用于多账户设置）\n- **fail** - 立即失败并给出明确错误信息（非常适合 CI\u002FCD）\n\n### 🎯 引导技能 - 自托管\n\n将 skill-seekers 打包为技能，以便在您的 AI 代理（Claude Code、Kimi、Codex 等）中使用：\n\n```bash\n# 生成技能\n.\u002Fscripts\u002Fbootstrap_skill.sh\n\n# 安装到 Claude Code\ncp -r output\u002Fskill-seekers ~\u002F.claude\u002Fskills\u002F\n```\n\n**您将获得：**\n- ✅ **完整的技能文档** - 所有 CLI 命令和使用模式\n- ✅ **CLI 命令参考** - 记录了每个工具及其选项\n- ✅ **快速入门示例** - 常见工作流程和最佳实践\n- ✅ **自动生成的 API 文档** - 包括代码分析、模式和示例\n\n### 🔐 私有配置仓库\n- ✅ **基于 Git 的配置源** - 从私有\u002F团队 Git 仓库获取配置\n- ✅ **多源管理** - 注册不限数量的 GitHub、GitLab、Bitbucket 仓库\n- ✅ **团队协作** - 在 3-5 人团队内共享自定义配置\n- ✅ **企业支持** - 可扩展至 500 多名开发者，并提供优先级解决机制\n- ✅ **安全认证** - 使用环境变量令牌（GITHUB_TOKEN、GITLAB_TOKEN）\n- ✅ **智能缓存** - 克隆一次，自动拉取更新\n- ✅ **离线模式** - 即使离线也能使用缓存配置\n\n### 🤖 代码库分析（C3.x）\n\n**C3.4：AI 增强的配置模式提取**\n- ✅ **9 种配置格式** - JSON、YAML、TOML、ENV、INI、Python、JavaScript、Dockerfile、Docker Compose\n- ✅ **7 种模式类型** - 数据库、API、日志、缓存、邮件、认证、服务器配置\n- ✅ **AI 增强** - 可选双模 AI 分析（API + LOCAL）\n  - 解释每种配置的作用\n  - 提供建议和改进建议\n  - **安全分析** - 查找硬编码的秘密和暴露的凭据\n- ✅ **自动文档化** - 生成所有配置的 JSON + Markdown 文档\n- ✅ **MCP 集成** - `extract_config_patterns` 工具支持增强功能\n\n**C3.3：AI 增强的教程指南**\n- ✅ **全面的 AI 增强** - 将基础指南转化为专业教程\n- ✅ **5 项自动改进** - 步骤描述、故障排除、先决条件、后续步骤、使用场景\n- ✅ **双模支持** - API 模式（Claude API）或 LOCAL 模式（Claude Code CLI）\n- ✅ **LOCAL 模式无需 API 费用** - 使用您的 Claude Code Max 方案即可免费增强\n- ✅ **质量飞跃** - 75 行模板 → 500+ 行综合指南\n\n**使用方法：**\n```bash\n# 快速分析（1-2 分钟，仅基础功能）\nskill-seekers analyze --directory tests\u002F --quick\n\n# 全面分析（20-60 分钟，所有功能）\nskill-seekers analyze --directory tests\u002F --comprehensive\n\n# 含 AI 增强\nskill-seekers analyze --directory tests\u002F --enhance\n```\n\n**完整文档**：[docs\u002FHOW_TO_GUIDES.md](docs\u002FHOW_TO_GUIDES.md#ai-enhancement-new)\n\n### 🔄 增强工作流预设\n\n可重用的 YAML 定义增强流水线，控制 AI 如何将您的原始文档转换为完善的技能。\n\n- ✅ **5 个捆绑预设** — `default`、`minimal`、`security-focus`、`architecture-comprehensive`、`api-documentation`\n- ✅ **用户自定义预设** — 将自定义工作流添加到 `~\u002F.config\u002Fskill-seekers\u002Fworkflows\u002F`\n- ✅ **多工作流** — 在一个命令中串联两个或多个工作流\n- ✅ **完全管理的 CLI** — 列出、检查、复制、添加、删除和验证工作流\n\n```bash\n# 应用单个工作流\nskill-seekers create .\u002Fmy-project --enhance-workflow security-focus\n\n# 串联多个工作流（按顺序应用）\nskill-seekers create .\u002Fmy-project \\\n  --enhance-workflow security-focus \\\n  --enhance-workflow minimal\n\n# 管理预设\nskill-seekers workflows list                          # 列出所有（捆绑 + 用户）\nskill-seekers workflows show security-focus           # 打印 YAML 内容\nskill-seekers workflows copy security-focus           # 复制到用户目录以便编辑\nskill-seekers workflows add .\u002Fmy-workflow.yaml        # 安装自定义预设\nskill-seekers workflows remove my-workflow            # 删除用户预设\nskill-seekers workflows validate security-focus       # 验证预设结构\n\n# 一次性复制多个\nskill-seekers workflows copy security-focus minimal api-documentation\n\n# 一次性添加多个文件\nskill-seekers workflows add .\u002Fwf-a.yaml .\u002Fwf-b.yaml\n\n# 一次性删除多个\nskill-seekers workflows remove my-wf-a my-wf-b\n```\n\n**YAML 预设格式：**\n```yaml\nname: security-focus\ndescription: \"以安全为重点的审查：漏洞、认证、数据处理\"\nversion: \"1.0\"\nstages:\n  - name: vulnerabilities\n    type: custom\n    prompt: \"审查 OWASP 十大常见安全漏洞...\"\n  - name: auth-review\n    type: custom\n    prompt: \"检查身份验证和授权模式...\"\n    uses_history: true\n```\n\n### ⚡ 性能与规模\n- ✅ **异步模式** - 使用 async\u002Fawait 可使抓取速度提升 2-3 倍（使用 `--async` 标志）\n- ✅ **大型文档支持** - 通过智能拆分，可处理 1 万至 4 万页以上的文档\n- ✅ **路由器\u002F集线器技能** - 智能路由至专业子技能\n- ✅ **并行抓取** - 同时处理多个技能\n- ✅ **断点续传** - 长时间抓取任务不会丢失进度\n- ✅ **缓存系统** - 抓取一次，即可立即重建\n\n### 🤖 代理无关的技能生成\n- ✅ **多代理支持** - 可为 Claude、Kimi、Codex、Copilot、OpenCode 或任何自定义代理生成技能，只需使用 `--agent` 标志\n- ✅ **自定义代理命令** - 使用 `--agent-cmd` 指定用于增强的自定义代理 CLI 命令\n- ✅ **通用标志** - `--agent` 和 `--agent-cmd` 适用于所有命令（创建、抓取、GitHub、PDF 等）\n\n### 📦 市场管道\n- ✅ **发布到市场** - 将技能发布到 Claude Code 插件市场仓库\n- ✅ **端到端管道** - 从文档源到已发布的市场条目\n\n### ✅ 质量保证\n- ✅ **全面测试** - 2,540 多项测试，覆盖全面\n\n---\n\n## 📦 安装\n\n```bash\n# 基本安装（文档抓取、GitHub 分析、PDF、打包）\npip install skill-seekers\n\n# 包含所有 LLM 平台支持\npip install skill-seekers[all-llms]\n\n# 包含 MCP 服务器\npip install skill-seekers[mcp]\n\n# 全部功能\npip install skill-seekers[all]\n```\n\n**需要帮助选择吗？** 运行设置向导：\n```bash\nskill-seekers-setup\n```\n\n### 安装选项\n\n| 安装 | 功能 |\n|---------|----------|\n| `pip install skill-seekers` | 抓取、GitHub 分析、PDF、所有平台 |\n| `pip install skill-seekers[gemini]` | + Google Gemini 支持 |\n| `pip install skill-seekers[openai]` | + OpenAI ChatGPT 支持 |\n| `pip install skill-seekers[all-llms]` | + 所有 LLM 平台 |\n| `pip install skill-seekers[mcp]` | + MCP 服务器，适用于 Claude Code、Cursor 等 |\n| `pip install skill-seekers[video]` | + YouTube\u002FVimeo 字幕及元数据提取 |\n| `pip install skill-seekers[video-full]` | + Whisper 转录及视觉帧提取 |\n| `pip install skill-seekers[jupyter]` | + Jupyter Notebook 支持 |\n| `pip install skill-seekers[pptx]` | + PowerPoint 支持 |\n| `pip install skill-seekers[confluence]` | + Confluence Wiki 支持 |\n| `pip install skill-seekers[notion]` | + Notion 页面支持 |\n| `pip install skill-seekers[rss]` | + RSS\u002FAtom 订阅支持 |\n| `pip install skill-seekers[chat]` | + Slack\u002FDiscord 聊天记录导出支持 |\n| `pip install skill-seekers[asciidoc]` | + AsciiDoc 文档支持 |\n| `pip install skill-seekers[all]` | 启用所有功能 |\n\n> **视频视觉依赖（GPU 感知）：** 安装 `skill-seekers[video-full]` 后，运行\n> `skill-seekers video --setup` 自动检测您的 GPU 并安装正确的 PyTorch\n> 版本 + easyocr。这是安装视觉提取依赖的推荐方式。\n\n---\n\n## 🚀 一键安装流程\n\n**从配置到上传技能的最快方式——完全自动化：**\n\n```bash\n# 从官方配置安装 React 技能（自动上传至 Claude）\nskill-seekers install --config react\n\n# 从本地配置文件安装\nskill-seekers install --config configs\u002Fcustom.json\n\n# 不上传（仅打包）\nskill-seekers install --config django --no-upload\n\n# 预览工作流而不执行\nskill-seekers install --config react --dry-run\n```\n\n**耗时：** 总共 20-45 分钟 | **质量：** 生产就绪（9\u002F10）| **成本：** 免费\n\n**执行阶段：**\n```\n📥 第 1 阶段：获取配置（如果提供了配置名称）\n📖 第 2 阶段：抓取文档\n✨ 第 3 阶段：AI 增强（强制性——不可跳过）\n📦 第 4 阶段：打包技能\n☁️ 第 5 阶段：上传至 Claude（可选，需 API 密钥）\n```\n\n**要求：**\n- ANTHROPIC_API_KEY 环境变量（用于自动上传）\n- Claude Code Max 方案（用于本地 AI 增强），或使用 `--agent` 选择其他 AI 代理\n\n---\n\n## 📊 功能矩阵\n\nSkill Seekers 支持 **12 个 LLM 平台**、**17 种来源类型**，并在所有目标上实现功能对等。\n\n**平台：** Claude AI、Google Gemini、OpenAI ChatGPT、MiniMax AI、通用 Markdown、OpenCode、Kimi（Moonshot AI）、DeepSeek AI、Qwen（Alibaba）、OpenRouter、Together AI、Fireworks AI  \n**来源类型：** 文档网站、GitHub 仓库、PDF、Word (.docx)、EPUB、视频、本地代码库、Jupyter 笔记books、本地 HTML、OpenAPI\u002FSwagger、AsciiDoc、PowerPoint (.pptx)、RSS\u002FAtom 订阅、Man 页面、Confluence 维基、Notion 页面、Slack\u002FDiscord 聊天记录导出\n\n详细平台和功能支持，请参阅 [完整功能矩阵](docs\u002FFEATURE_MATRIX.md)。\n\n### 快速平台比较\n\n| 功能 | Claude | Gemini | OpenAI | MiniMax | Markdown |\n|---------|--------|--------|--------|--------|----------|\n| 格式 | ZIP + YAML | tar.gz | ZIP + Vector | ZIP + Knowledge | ZIP |\n| 上传 | ✅ API | ✅ API | ✅ API | ✅ API | ❌ 手动 |\n| 增强 | ✅ Sonnet 4 | ✅ 2.0 Flash | ✅ GPT-4o | ✅ M2.7 | ❌ 无 |\n| 所有技能模式 | ✅ | ✅ | ✅ | ✅ | ✅ |\n\n---\n\n## 使用示例\n\n### 文档抓取\n\n```bash\n# 抓取文档网站\nskill-seekers scrape --config configs\u002Freact.json\n\n# 无需配置的快速抓取\nskill-seekers scrape --url https:\u002F\u002Freact.dev --name react\n\n# 使用异步模式（速度提升3倍）\nskill-seekers scrape --config configs\u002Fgodot.json --async --workers 8\n\n# 使用特定的AI智能体进行增强\nskill-seekers scrape --config configs\u002Freact.json --agent kimi\n```\n\n### PDF 提取\n\n```bash\n# 基本PDF提取\nskill-seekers pdf --pdf docs\u002Fmanual.pdf --name myskill\n\n# 高级功能\nskill-seekers pdf --pdf docs\u002Fmanual.pdf --name myskill \\\n    --extract-tables \\        # 提取表格\n    --parallel \\              # 快速并行处理\n    --workers 8               # 使用8个CPU核心\n\n# 扫描版PDF（需安装：pip install pytesseract Pillow）\nskill-seekers pdf --pdf docs\u002Fscanned.pdf --name myskill --ocr\n```\n\n### 视频提取\n\n```bash\n# 安装视频支持\npip install skill-seekers[video]        # 提供字幕和元数据\npip install skill-seekers[video-full]   # 包含Whisper和视觉帧提取\n\n# 自动检测GPU并安装视觉依赖项（PyTorch + easyocr）\nskill-seekers video --setup\n\n# 从YouTube视频中提取内容\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=dQw4w9WgXcQ --name mytutorial\n\n# 从YouTube播放列表中提取内容\nskill-seekers video --playlist https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=... --name myplaylist\n\n# 从本地视频文件中提取内容\nskill-seekers video --video-file recording.mp4 --name myrecording\n\n# 带视觉帧分析的提取（需安装video-full依赖）\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --name mytutorial --visual\n\n# 使用AI增强（清理OCR文本并生成精美的SKILL.md）\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --visual --enhance-level 2\n\n# 截取视频的特定片段（支持秒数、MM:SS、HH:MM:SS格式）\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --start-time 1:30 --end-time 5:00\n\n# 对低置信度OCR帧使用Vision API（需设置ANTHROPIC_API_KEY）\nskill-seekers video --url https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=... --visual --vision-ocr\n\n# 从先前提取的数据重建技能（跳过下载步骤）\nskill-seekers video --from-json output\u002Fmytutorial\u002Fvideo_data\u002Fextracted_data.json --name mytutorial\n```\n\n> **完整指南：** 请参阅 [docs\u002FVIDEO_GUIDE.md](docs\u002FVIDEO_GUIDE.md)，了解完整的CLI参考、视觉流程细节、AI增强选项以及故障排除方法。\n\n### GitHub 仓库分析\n\n```bash\n# 基本仓库抓取\nskill-seekers github --repo facebook\u002Freact\n\n# 使用认证（提高速率限制）\nexport GITHUB_TOKEN=ghp_your_token_here\nskill-seekers github --repo facebook\u002Freact\n\n# 自定义包含内容\nskill-seekers github --repo django\u002Fdjango \\\n    --include-issues \\        # 提取GitHub Issues\n    --max-issues 100 \\        # 限制Issue数量\n    --include-changelog       # 提取CHANGELOG.md\n```\n\n### 统一多源抓取\n\n**将文档、GitHub和PDF整合为一个统一的技能，并自动检测冲突：**\n\n```bash\n# 使用现有的统一配置\nskill-seekers unified --config configs\u002Freact_unified.json\nskill-seekers unified --config configs\u002Fdjango_unified.json\n\n# 或者创建统一配置\ncat > configs\u002Fmyframework_unified.json \u003C\u003C 'EOF'\n{\n  \"name\": \"myframework\",\n  \"merge_mode\": \"rule-based\",\n  \"sources\": [\n    {\n      \"type\": \"documentation\",\n      \"base_url\": \"https:\u002F\u002Fdocs.myframework.com\u002F\",\n      \"max_pages\": 200\n    },\n    {\n      \"type\": \"github\",\n      \"repo\": \"owner\u002Fmyframework\",\n      \"code_analysis_depth\": \"surface\"\n    }\n  ]\n}\nEOF\n\nskill-seekers unified --config configs\u002Fmyframework_unified.json\n```\n\n**冲突检测会自动识别：**\n- 🔴 **代码中缺失**（高优先级）：文档中有但代码中未实现\n- 🟡 **文档中缺失**（中优先级）：代码中已实现但文档未记录\n- ⚠️ **签名不匹配**：参数或类型不同\n- ℹ️ **描述不匹配**：解释内容不一致\n\n**完整指南：** 请参阅 [docs\u002FUNIFIED_SCRAPING.md](docs\u002FUNIFIED_SCRAPING.md) 获取完整说明。\n\n### 私有配置仓库\n\n**通过私有Git仓库在团队间共享自定义配置：**\n\n```bash\n# 选项1：使用MCP工具（推荐）\n# 注册团队的私有仓库\nadd_config_source(\n    name=\"team\",\n    git_url=\"https:\u002F\u002Fgithub.com\u002Fmycompany\u002Fskill-configs.git\",\n    token_env=\"GITHUB_TOKEN\"\n)\n\n# 从团队仓库获取配置\nfetch_config(source=\"team\", config_name=\"internal-api\")\n```\n\n**支持的平台：**\n- GitHub (`GITHUB_TOKEN`)、GitLab (`GITLAB_TOKEN`)、Gitea (`GITEA_TOKEN`)、Bitbucket (`BITBUCKET_TOKEN`)\n\n**完整指南：** 请参阅 [docs\u002FGIT_CONFIG_SOURCES.md](docs\u002FGIT_CONFIG_SOURCES.md) 获取完整说明。\n\n## 工作原理\n\n```mermaid\ngraph LR\n    A[文档网站] --> B[Skill Seekers]\n    B --> C[抓取器]\n    B --> D[AI增强]\n    B --> E[打包器]\n    C --> F[整理后的参考资料]\n    D --> F\n    F --> E\n    E --> G[AI技能.zip]\n    G --> H[上传至AI平台]\n```\n\n0. **检测llms.txt** - 首先检查llms-full.txt、llms.txt、llms-small.txt（属于Smart SPA Discovery的一部分）\n1. **抓取**：提取文档中的所有页面\n2. **分类**：将内容组织成主题（API、指南、教程等）\n3. **增强**：AI分析文档并创建包含示例的全面SKILL.md文件（可通过`--agent`参数支持多种智能体）\n4. **打包**：将所有内容打包成适合平台使用的`.zip`文件\n\n## 架构\n\n系统由**8个核心模块**和**5个工具模块**组成（总计约200个类）：\n\n![包概览](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_9101963a8489.png)\n\n| 模块 | 目的 | 关键类 |\n|--------|---------|-------------|\n| **CLICore** | Git风格的命令分发器 | `CLIDispatcher`、`SourceDetector`、`CreateCommand` |\n| **Scrapers** | 17种来源类型的提取器 | `DocToSkillConverter`、`GitHubScraper`、`UnifiedScraper` |\n| **Adaptors** | 20多种输出平台格式 | `SkillAdaptor`（ABC）、`ClaudeAdaptor`、`LangChainAdaptor` |\n| **Analysis** | C3.x代码库分析流水线 | `UnifiedCodebaseAnalyzer`、`PatternRecognizer`、10种GoF检测器 |\n| **Enhancement** | 基于AI的技能改进，通过`AgentClient`实现 | `AgentClient`、`AIEnhancer`、`UnifiedEnhancer`、`WorkflowEngine` |\n| **Packaging** | 打包、上传、安装技能 | `PackageSkill`、`InstallAgent` |\n| **MCP** | FastMCP服务器（40种工具） | `SkillSeekerMCPServer`、10个工具模块 |\n| **Sync** | 文档变更检测 | `ChangeDetector`、`SyncMonitor`、`Notifier` |\n\n工具模块：**解析器**（28种CLI解析器）、**存储**（S3\u002FGCS\u002FAzure）、**嵌入**（多提供商向量）、**基准测试**（性能）、**实用工具**（16种共享助手）。\n\n完整UML图：**[docs\u002FUML_ARCHITECTURE.md](docs\u002FUML_ARCHITECTURE.md)** | StarUML项目：`docs\u002FUML\u002Fskill_seekers.mdj` | HTML API参考：`docs\u002FUML\u002Fhtml\u002F`\n\n## 📋 前置条件\n\n**在开始之前，请确保您已具备以下内容：**\n\n1. **Python 3.10 或更高版本** - [下载](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002F) | 检查命令：`python3 --version`\n2. **Git** - [下载](https:\u002F\u002Fgit-scm.com\u002F) | 检查命令：`git --version`\n3. **15-30 分钟** 用于首次设置\n\n**首次使用？** → **[从这里开始：防弹快速入门指南](BULLETPROOF_QUICKSTART.md)** 🎯\n\n---\n\n## 📤 将技能上传至 Claude\n\n当您的技能打包完成后，需要将其上传至 Claude：\n\n### 选项 1：自动上传（基于 API）\n\n```bash\n# 设置您的 API 密钥（仅需一次）\nexport ANTHROPIC_API_KEY=sk-ant-...\n\n# 自动打包并上传\nskill-seekers package output\u002Freact\u002F --upload\n\n# 或者上传现有的 .zip 文件\nskill-seekers upload output\u002Freact.zip\n```\n\n### 选项 2：手动上传（无需 API 密钥）\n\n```bash\n# 打包技能\nskill-seekers package output\u002Freact\u002F\n# → 生成 output\u002Freact.zip\n\n# 然后手动上传：\n# - 访问 https:\u002F\u002Fclaude.ai\u002Fskills\n# - 点击“上传技能”\n# - 选择 output\u002Freact.zip\n```\n\n### 选项 3：MCP（Claude Code）\n\n```\n在 Claude Code 中，只需询问：\n“打包并上传 React 技能”\n```\n\n---\n\n## 🤖 安装到 AI 代理\n\nSkill Seekers 可以自动将技能安装到 18 种 AI 编程代理中。\n\n```bash\n# 安装到特定代理\nskill-seekers install-agent output\u002Freact\u002F --agent cursor\n\n# 一次性安装到所有代理\nskill-seekers install-agent output\u002Freact\u002F --agent all\n\n# 预览而不安装\nskill-seekers install-agent output\u002Freact\u002F --agent cursor --dry-run\n```\n\n### 支持的代理\n\n| 代理 | 路径 | 类型 |\n|-------|------|------|\n| **Claude Code** | `~\u002F.claude\u002Fskills\u002F` | 全局 |\n| **Cursor** | `.cursor\u002Fskills\u002F` | 项目 |\n| **VS Code \u002F Copilot** | `.github\u002Fskills\u002F` | 项目 |\n| **Amp** | `~\u002F.amp\u002Fskills\u002F` | 全局 |\n| **Goose** | `~\u002F.config\u002Fgoose\u002Fskills\u002F` | 全局 |\n| **OpenCode** | `~\u002F.opencode\u002Fskills\u002F` | 全局 |\n| **Windsurf** | `~\u002F.windsurf\u002Fskills\u002F` | 全局 |\n| **Roo Code** | `.roo\u002Fskills\u002F` | 项目 |\n| **Cline** | `.cline\u002Fskills\u002F` | 项目 |\n| **Aider** | `~\u002F.aider\u002Fskills\u002F` | 全局 |\n| **Bolt** | `.bolt\u002Fskills\u002F` | 项目 |\n| **Kilo Code** | `.kilo\u002Fskills\u002F` | 项目 |\n| **Continue** | `~\u002F.continue\u002Fskills\u002F` | 全局 |\n| **Kimi Code** | `~\u002F.kimi\u002Fskills\u002F` | 全局 |\n\n---\n\n## 🔌 MCP 集成（26 种工具）\n\nSkill Seekers 提供了一个 MCP 服务器，可用于 Claude Code、Cursor、Windsurf、VS Code + Cline 或 IntelliJ IDEA。\n\n```bash\n# stdio 模式（Claude Code、VS Code + Cline）\npython -m skill_seekers.mcp.server_fastmcp\n\n# HTTP 模式（Cursor、Windsurf、IntelliJ）\npython -m skill_seekers.mcp.server_fastmcp --transport http --port 8765\n\n# 一次性自动配置所有代理\n.\u002Fsetup_mcp.sh\n```\n\n**所有 26 种工具可用：**\n- **核心（9）：** `list_configs`、`generate_config`、`validate_config`、`estimate_pages`、`scrape_docs`、`package_skill`、`upload_skill`、`enhance_skill`、`install_skill`\n- **扩展（10）：** `scrape_github`、`scrape_pdf`、`unified_scrape`、`merge_sources`、`detect_conflicts`、`add_config_source`、`fetch_config`、`list_config_sources`、`remove_config_source`、`split_config`\n- **向量数据库（4）：** `export_to_chroma`、`export_to_weaviate`、`export_to_faiss`、`export_to_qdrant`\n- **云服务（3）：** `cloud_upload`、`cloud_download`、`cloud_list`\n\n**完整指南：** [docs\u002FMCP_SETUP.md](docs\u002FMCP_SETUP.md)\n\n---\n\n## ⚙️ 配置\n\n### 可用预设（24+）\n\n```bash\n# 列出所有预设\nskill-seekers list-configs\n```\n\n| 类别 | 预设 |\n|----------|---------|\n| **Web 框架** | `react`、`vue`、`angular`、`svelte`、`nextjs` |\n| **Python** | `django`、`flask`、`fastapi`、`sqlalchemy`、`pytest` |\n| **游戏开发** | `godot`、`pygame`、`unity` |\n| **工具与 DevOps** | `docker`、`kubernetes`、`terraform`、`ansible` |\n| **统一版（文档 + GitHub）** | `react-unified`、`vue-unified`、`nextjs-unified` 等等\n\n### 创建您自己的配置\n\n```bash\n# 选项 1：交互式\nskill-seekers scrape --interactive\n\n# 选项 2：复制并编辑预设\ncp configs\u002Freact.json configs\u002Fmyframework.json\nnano configs\u002Fmyframework.json\nskill-seekers scrape --config configs\u002Fmyframework.json\n```\n\n### 配置文件结构\n\n```json\n{\n  \"name\": \"myframework\",\n  \"description\": \"何时使用此技能\",\n  \"base_url\": \"https:\u002F\u002Fdocs.myframework.com\u002F\",\n  \"selectors\": {\n    \"main_content\": \"article\",\n    \"title\": \"h1\",\n    \"code_blocks\": \"pre code\"\n  },\n  \"url_patterns\": {\n    \"include\": [\"\u002Fdocs\", \"\u002Fguide\"],\n    \"exclude\": [\"\u002Fblog\", \"\u002Fabout\"]\n  },\n  \"categories\": {\n    \"getting_started\": [\"intro\", \"quickstart\"],\n    \"api\": [\"api\", \"reference\"]\n  },\n  \"rate_limit\": 0.5,\n  \"max_pages\": 500\n}\n```\n\n### 配置存储位置\n\n该工具会按以下顺序搜索配置：\n1. 提供的确切路径\n2. `.\u002Fconfigs\u002F`（当前目录）\n3. `~\u002F.config\u002Fskill-seekers\u002Fconfigs\u002F`（用户配置目录）\n4. SkillSeekersWeb.com API（预设配置）\n\n---\n\n## 📊 生成的内容\n\n```\noutput\u002F\n├── godot_data\u002F              # 抓取的原始数据\n│   ├── pages\u002F              # JSON 文件（每页一个）\n│   └── summary.json        # 概述\n│\n└── godot\u002F                   # 生成的技能\n    ├── SKILL.md            # 使用真实示例增强\n    ├── references\u002F         # 分类后的文档\n    │   ├── index.md\n    │   ├── getting_started.md\n    │   ├── scripting.md\n    │   └── ...\n    ├── scripts\u002F            # 空的（可自行添加）\n    └── assets\u002F             # 空的（可自行添加）\n```\n\n---\n\n## 🐛 故障排除\n\n### 没有提取到内容？\n- 检查您的 `main_content` 选择器\n- 尝试：`article`、`main`、`div[role=\"main\"]`\n\n### 数据存在但不被使用？\n```bash\n# 强制重新抓取\nrm -rf output\u002Fmyframework_data\u002F\nskill-seekers scrape --config configs\u002Fmyframework.json\n```\n\n### 分类不够好？\n编辑配置中的 `categories` 部分，使用更合适的关键词。\n\n### 想更新文档？\n```bash\n# 删除旧数据并重新抓取\nrm -rf output\u002Fgodot_data\u002F\nskill-seekers scrape --config configs\u002Fgodot.json\n```\n\n### 增强功能不起作用？\n```bash\n# 检查是否设置了 API 密钥\necho $ANTHROPIC_API_KEY\n\n# 尝试使用 LOCAL 模式（使用 Claude Code Max，无需 API 密钥）\nskill-seekers enhance output\u002Freact\u002F --mode LOCAL\n\n# 监控后台增强状态\nskill-seekers enhance-status output\u002Freact\u002F --watch\n```\n\n### GitHub 速率限制问题？\n```bash\n# 设置 GitHub 令牌（5000 请求\u002F小时 vs 匿名用户的 60 请求\u002F小时）\nexport GITHUB_TOKEN=ghp_your_token_here\n\n# 或者配置多个个人资料\nskill-seekers config --github\n```\n\n---\n\n## 📈 性能\n\n| 任务 | 时间 | 备注 |\n|------|------|-------|\n| 抓取（同步） | 15-45 分钟 | 仅首次执行，基于线程 |\n| 抓取（异步） | 5-15 分钟 | 使用 `--async` 标志时速度提升 2-3 倍 |\n| 构建 | 1-3 分钟 | 从缓存快速重建 |\n| 重新构建 | \u003C1 分钟 | 使用 `--skip-scrape` 标志 |\n| 增强（LOCAL） | 30-60 秒 | 使用 Claude Code Max |\n| 增强（API） | 20-40 秒 | 需要 API 密钥 |\n| 视频（转录本） | 1-3 分钟 | 仅 YouTube\u002F本地视频的转录本 |\n| 视频（视觉内容） | 5-15 分钟 | + OCR 帧提取 |\n| 打包 | 5-10 秒 | 最终生成 .zip 文件 |\n\n---\n\n## 📚 文档\n\n### 入门\n- **[BULLETPROOF_QUICKSTART.md](BULLETPROOF_QUICKSTART.md)** - 🎯 如果你是新手，请从这里开始！\n- **[QUICKSTART.md](QUICKSTART.md)** - 高级用户的快速入门\n- **[TROUBLESHOOTING.md](TROUBLESHOOTING.md)** - 常见问题及解决方案\n- **[docs\u002FQUICK_REFERENCE.md](docs\u002FQUICK_REFERENCE.md)** - 一页纸的速查表\n\n### 架构\n- **[docs\u002FUML_ARCHITECTURE.md](docs\u002FUML_ARCHITECTURE.md)** - 包含14张图的UML架构概览\n- **[docs\u002FUML\u002Fexports\u002F](docs\u002FUML\u002Fexports\u002F)** - PNG格式的图表导出文件（包概述 + 13张类图）\n- **[docs\u002FUML\u002Fhtml\u002F](docs\u002FUML\u002Fhtml\u002Findex.html\u002Findex.html)** - 完整的HTML API参考文档（所有类、操作和属性）\n- **[docs\u002FUML\u002Fskill_seekers.mdj](docs\u002FUML\u002Fskill_seekers.mdj)** - StarUML项目文件（可用[StarUML](https:\u002F\u002Fstaruml.io\u002F)打开）\n\n### 指南\n- **[docs\u002FLARGE_DOCUMENTATION.md](docs\u002FLARGE_DOCUMENTATION.md)** - 处理1万至4万页以上的文档\n- **[ASYNC_SUPPORT.md](ASYNC_SUPPORT.md)** - 异步模式指南（爬取速度提升2至3倍）\n- **[docs\u002FENHANCEMENT_MODES.md](docs\u002FENHANCEMENT_MODES.md)** - AI增强模式指南\n- **[docs\u002FMCP_SETUP.md](docs\u002FMCP_SETUP.md)** - MCP集成设置\n- **[docs\u002FUNIFIED_SCRAPING.md](docs\u002FUNIFIED_SCRAPING.md)** - 多源数据抓取\n- **[docs\u002FVIDEO_GUIDE.md](docs\u002FVIDEO_GUIDE.md)** - 视频提取指南\n\n### 集成指南\n- **[docs\u002Fintegrations\u002FLANGCHAIN.md](docs\u002Fintegrations\u002FLANGCHAIN.md)** - LangChain RAG\n- **[docs\u002Fintegrations\u002FCURSOR.md](docs\u002Fintegrations\u002FCURSOR.md)** - Cursor IDE\n- **[docs\u002Fintegrations\u002FWINDSURF.md](docs\u002Fintegrations\u002FWINDSURF.md)** - Windsurf IDE\n- **[docs\u002Fintegrations\u002FCLINE.md](docs\u002Fintegrations\u002FCLINE.md)** - Cline（VS Code）\n- **[docs\u002Fintegrations\u002FRAG_PIPELINES.md](docs\u002Fintegrations\u002FRAG_PIPELINES.md)** - 所有RAG管道\n\n---\n\n## 📝 许可证\n\nMIT许可证 - 详情请参阅[LICENSE](LICENSE)文件\n\n---\n\n祝你技能提升愉快！🚀\n\n---\n\n## 🔒 安全性\n\n[![MseeP.ai安全评估徽章](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_readme_f0c1c269426f.png)](https:\u002F\u002Fmseep.ai\u002Fapp\u002Fyusufkaraaslan-skill-seekers)","# Skill Seekers 快速上手指南\n\nSkill Seekers 是专为 AI 系统打造的数据层工具。它能将文档网站、GitHub 仓库、PDF、视频、Notebook 等 17+ 种来源转化为结构化的知识资产，直接用于构建 Claude Skills、RAG 管道（LangChain, LlamaIndex）或 AI 编程助手（Cursor, Windsurf）。\n\n## 环境准备\n\n在开始之前，请确保您的开发环境满足以下要求：\n\n*   **操作系统**：Linux, macOS 或 Windows (WSL 推荐)\n*   **Python 版本**：Python 3.10 或更高版本\n*   **包管理器**：`pip` (建议配置国内镜像源以加速下载)\n*   **可选依赖**：若需处理视频内容，需额外安装 GPU 相关的视觉依赖库。\n\n> **💡 国内加速建议**\n> 建议使用清华或阿里镜像源安装 Python 包，以提升下载速度：\n> ```bash\n> pip install -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple \u003Cpackage_name>\n> ```\n\n## 安装步骤\n\n通过 PyPI 一键安装核心工具：\n\n```bash\n# 使用默认源安装\npip install skill-seekers\n\n# 【推荐】使用清华镜像源安装（国内用户）\npip install -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple skill-seekers\n```\n\n**验证安装：**\n```bash\nskill-seekers --version\n```\n\n> **注意**：如果您需要处理视频源（YouTube 或本地视频），请先运行以下命令自动安装所需的视觉依赖：\n> ```bash\n> skill-seekers video --setup\n> ```\n\n## 基本使用\n\n只需三个命令即可完成从数据源到 AI 可用资产的转换。\n\n### 1. 创建技能资产 (Create)\n从任意支持的来源（如官方文档网址、GitHub 仓库、本地文件）抓取并结构化数据。\n\n```bash\n# 示例：从 Django 官方文档创建技能\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F\n\n# 其他常用来源示例：\n# skill-seekers create facebook\u002Freact          # GitHub 仓库\n# skill-seekers create .\u002Fmy-project            # 本地项目目录\n# skill-seekers create manual.pdf              # PDF 文档\n```\n\n执行完成后，数据将被处理并保存到 `output\u002F` 目录中。\n\n### 2. 打包导出 (Package)\n将处理好的资产打包为目标 AI 平台所需的格式。\n\n```bash\n# 示例：打包为 Claude AI Skill (生成 ZIP + YAML)\nskill-seekers package output\u002Fdjango --target claude\n\n# 示例：打包为 LangChain 文档格式\nskill-seekers package output\u002Fdjango --target langchain\n\n# 示例：打包为 Cursor 编辑器规则 (.cursorrules)\nskill-seekers package output\u002Fdjango --target cursor\n```\n\n### 3. 完成\n命令执行成功后，您将在 `output\u002F` 目录下获得对应的文件（如 `output\u002Fdjango-claude.zip`），可直接上传至相应的 AI 平台或集成到您的 RAG  pipeline 中。\n\n---\n\n### 进阶：指定增强模型\n默认使用 Claude 进行内容增强，您也可以指定其他模型：\n\n```bash\n# 使用 Kimi 进行增强\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F --agent kimi\n\n# 使用自定义命令调用本地 Agent\nskill-seekers create https:\u002F\u002Fdocs.django.com\u002F --agent-cmd \"my-custom-agent run\"\n```","某初创公司的后端团队急需将分散在官方文档站、GitHub 私有仓库及内部 PDF 规范中的微服务架构知识，快速转化为 Claude AI 可理解的专属技能，以辅助新入职工程师进行代码开发。\n\n### 没有 Skill_Seekers 时\n- **人工整理耗时极长**：开发人员需手动复制粘贴网页内容、下载 PDF 并清洗格式，耗费数天才能拼凑出一份完整的知识库。\n- **信息冲突难以察觉**：不同来源的文档（如旧版 PDF 与新版 GitHub README）存在逻辑矛盾，人工核对极易遗漏，导致 AI 学习到错误指令。\n- **知识更新滞后**：一旦上游代码或文档变更，重新同步数据需要重复繁琐的手工流程，AI 助手往往基于过时的信息进行回答。\n- **非结构化数据难利用**：视频演示、Jupyter Notebook 等多模态资料无法直接被 AI 读取，大量高价值技术细节被闲置。\n\n### 使用 Skill_Seekers 后\n- **分钟级自动构建**：只需配置源地址，Skill_Seekers 即可自动抓取网站、仓库和 PDF，并在几分钟内将其转换为结构化的 AI 技能资产。\n- **智能冲突检测**：工具内置算法自动识别并标记不同来源间的内容冲突，确保输入给 AI 的知识库逻辑一致且准确。\n- **实时同步机制**：结合 CI\u002FCD 流水线，当源代码或文档更新时，Skill_Seekers 自动触发重新索引，保证 AI 始终掌握最新技术规范。\n- **多源异构支持**：轻松处理视频、Wiki、Notebook 等 10+ 种复杂数据源，将原本沉睡的非结构化资料全部转化为可调用的 AI 能力。\n\nSkill_Seekers 通过将杂乱的多源技术文档瞬间转化为高质量、无冲突的结构化知识，让企业构建专属 AI 助手的时间从数天缩短至数分钟。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fyusufkaraaslan_Skill_Seekers_86f7712e.png","yusufkaraaslan","yusyus","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fyusufkaraaslan_57db38ce.jpg",null,"yusufkaraaslan.yk@gmail.com","https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan",[83,87,91,95],{"name":84,"color":85,"percentage":86},"Python","#3572A5",98.9,{"name":88,"color":89,"percentage":90},"Shell","#89e051",1.1,{"name":92,"color":93,"percentage":94},"Dockerfile","#384d54",0,{"name":96,"color":97,"percentage":94},"Go Template","#00ADD8",12182,1220,"2026-04-03T05:19:11","MIT","Linux, macOS, Windows","非必需。仅在使用视频处理功能（skill-seekers[video]）时可能需要 GPU 加速，具体型号和显存未说明。","未说明",{"notes":106,"python":107,"dependencies":108},"该工具主要作为 CLI 和 MCP 服务器运行。基础安装仅需 Python 环境；若需处理视频源（YouTube 或本地视频），需安装额外的可选依赖 'skill-seekers[video]'，此时可能涉及 GPU 相关的视觉处理库。支持通过 Homebrew 在 macOS 上安装。","3.10+",[109,110],"skill-seekers (核心包)","skill-seekers[video] (可选，用于视频处理)",[14,15,43,45],[113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130],"ai-tools","automation","claude-ai","claude-skills","documentation","documentation-generator","python","web-scraping","mcp","mcp-server","github","ocr","pdf","ast-parser","code-analysis","conflict-detection","github-scraper","multi-source",7,"2026-03-27T02:49:30.150509","2026-04-06T07:16:04.967247",[135,140,145,150,154,158],{"id":136,"question_zh":137,"answer_zh":138,"source_url":139},11927,"为什么在 Windows PowerShell 中运行 'skill-seekers' 命令时提示 'not recognized'？","这通常是因为工具未正确安装或未添加到系统路径中。请确保你已激活虚拟环境（如果使用），或者使用 `pip install -e .` 在当前目录下进行可编辑安装。如果是按照指南操作，请检查是否成功执行了安装步骤。对于 Windows 用户，创建配置文件时不能使用 Linux 的 `cat > file \u003C\u003C 'EOF'` 语法，应使用 PowerShell 特有的语法：\n@\"\n{\n  \"name\": \"test-skill\",\n  \"base_url\": \"https:\u002F\u002Ftailwindcss.com\u002Fdocs\u002Finstallation\",\n  \"selectors\": {\n    \"main_content\": \"#content-wrapper\",\n    \"title\": \"h1, h2, h3\",\n    \"code_blocks\": \"pre code\"\n  },\n  \"max_pages\": 5\n}\n\"@ | Out-File -FilePath configs\u002Ftest.json -Encoding utf8","https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fissues\u002F261",{"id":141,"question_zh":142,"answer_zh":143,"source_url":144},11928,"为什么抓取显示 'Scraped N pages' 但最终提示 'No scraped data found'？","这是因为存在两个主要问题：\n1. **计数误导**：'Scraped N pages' 统计的是访问过的 URL 数量，而不是实际保存到磁盘的页面数。如果页面内容少于 50 个字符，保存函数会静默跳过，但计数器仍会增加。\n2. **SPA 网站内容为空**：像 Discord 文档这样的 React 单页应用（SPA），直接使用 `requests.get()` 获取的 HTML 仅包含 `\u003Cdiv id=\"root\">\u003C\u002Fdiv>` 和 JS 包，没有实际内容。必须使用能执行 JavaScript 的浏览器引擎才能抓取到内容。\n解决方案是更新到修复了页面计数逻辑的版本，并确认目标网站是否需要 JS 渲染。","https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fissues\u002F277",{"id":146,"question_zh":147,"answer_zh":148,"source_url":149},11929,"抓取过程中出现 'Invalid IPv6 URL' 错误怎么办？","该错误通常发生在 Python 3.14+ 版本中，因为新版 `urlparse()` 对 URL 格式校验更严格，遇到文档中不完整的 IPv6 占位符（如 `http:\u002F\u002F[fdaa:x:x...`）时会直接抛出异常。\n解决方法是升级到包含修复补丁的版本（如 development 分支或 v3.2.0 之后的版本）。修复后的 `sanitize_url()` 函数会捕获 `ValueError`，先对括号进行编码处理后再重试解析。你可以尝试运行：\npip install git+https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers.git@development","https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fissues\u002F284",{"id":151,"question_zh":152,"answer_zh":153,"source_url":144},11930,"为什么抓取到的 URL 会被错误地追加 '\u002Findex.html.md' 导致 404？","旧版本代码中的 `_convert_to_md_urls()` 函数会盲目地向所有非 `.md` 结尾的 URL 追加 `\u002Findex.html.md`，这种模式仅适用于 Docusaurus 构建的网站，对其他站点（如 Discord 文档）会导致大量 404 错误。\n已在后续版本中修复：\n1. 不再自动追加 `\u002Findex.html.md`，URL 保持原样（仅去除锚点并去重）。\n2. 修复了 `.md` 后缀的判断逻辑，改用 `urlparse().path.endswith(\".md\")` 避免误判。\n请升级到最新版本以解决此问题。",{"id":155,"question_zh":156,"answer_zh":157,"source_url":139},11931,"之前的 '--interactive' 模式命令为什么无法使用了？","维护者已在最新版本中移除了 `--interactive` 参数。如果你在使用旧教程或文档时看到该命令，请注意它已不再受支持。当前版本推荐使用配置文件（JSON 格式）来定义抓取规则（如 CSS 选择器、URL 模式等），而不是通过交互式命令行输入。请查阅最新的 README 或文档以获取基于配置文件的用法示例。",{"id":159,"question_zh":160,"answer_zh":161,"source_url":139},11932,"如何为特定网站配置正确的 CSS 选择器以避免抓取不到内容？","如果抓取后内容为空，通常是因为默认的选择器无法匹配目标网站的 HTML 结构。你需要在配置文件中明确指定 `selectors` 字段。例如，针对 Tailwind CSS 文档，配置如下：\n{\n  \"name\": \"test-skill\",\n  \"base_url\": \"https:\u002F\u002Ftailwindcss.com\u002Fdocs\u002Finstallation\",\n  \"selectors\": {\n    \"main_content\": \"#content-wrapper\",\n    \"title\": \"h1, h2, h3\",\n    \"code_blocks\": \"pre code\"\n  },\n  \"max_pages\": 5\n}\n你可以通过浏览器开发者工具查看目标页面的 HTML 结构，找到包含主要内容的容器 ID 或类名，并填入 `main_content` 字段。",[163,168,173,178,183,188,193,198,203,208,213,218,223,228,233,238,243,248,253,258],{"id":164,"version":165,"summary_zh":166,"released_at":167},62346,"v3.4.0","## v3.4.0 新增功能\n\n**主题：** 新增8个LLM平台适配器（总计12个）、7条新的CLI代理路径（总计18条）、OpenCode技能工具、SPA站点检测、8处错误修复，以及完整的UML架构文档。\n\n### 平台扩展：从5个增至12个LLM目标\n\n| 新平台 | 标志 | 基础 |\n|---|---|---|\n| OpenCode | `--target opencode` | 基于目录、双YAML |\n| Kimi | `--target kimi` | 兼容OpenAI |\n| DeepSeek | `--target deepseek` | 兼容OpenAI |\n| Qwen | `--target qwen` | 兼容OpenAI |\n| OpenRouter | `--target openrouter` | 兼容OpenAI |\n| Together AI | `--target together` | 兼容OpenAI |\n| Fireworks AI | `--target fireworks` | 兼容OpenAI |\n\n所有新平台均继承自一个共享的**兼容OpenAI的基础类**，以确保行为一致性。\n\n### 代理扩展：从11条增至18条安装路径\n\n新增代理：**roo**、**cline**、**aider**、**bolt**、**kilo**、**continue**、**kimi-code**\n\n### OpenCode技能工具\n\n- **技能拆分器** — 自动将大型文档拆分为专注的子技能，并配备路由功能\n- **双向转换器** — 支持在OpenCode与任何平台格式之间导入导出\n\n### 发布内容\n\n- **Smithery清单** (`smithery.yaml`)\n- **GitHub Actions模板**，用于自动化技能更新\n- **Claude代码插件**，支持斜杠命令\n\n### 错误修复\n\n- 在Python 3.14严格`urlparse`模式下，`sanitize_url()`会崩溃 (#284)\n- 盲目追加`\u002Findex.html.md`会导致非Docusaurus站点损坏 (#277)\n- 统一了爬虫临时配置格式 (#317)\n- Unicode箭头符号会导致Windows cp1252终端显示异常\n- 插件斜杠命令中包含CLI标志\n- MiniMax适配器改进 (#319)\n- **误导性的“已抓取N页”计数** — 现在显示为`(N保存，M跳过)` (#320)\n- **SPA站点检测** — 当站点需要JavaScript渲染时会发出警告 (#320, #321)\n\n### 文档\n\n- **完整UML架构** — 通过StarUML从源代码同步生成14张类图\n- StarUML HTML API参考导出\n- 生态系统部分链接所有Skill Seekers仓库\n- README和CONTRIBUTING中添加了架构引用\n- 将`Docs\u002F`合并至`docs\u002F`\n\n### 测试结果\n\n2929项通过，39项被跳过，无失败\n\n### 安装\u002F升级\n\n```bash\npip install --upgrade skill-seekers\n```\n\n**完整变更日志：** https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FCHANGELOG.md","2026-03-25T19:21:47",{"id":169,"version":170,"summary_zh":171,"released_at":172},62347,"v3.3.0","## [3.3.0] - 2026-03-16\n\n**主题:** 新增10种源类型（总计17种）、EPUB统一集成、sync-config命令、性能优化、12份README翻译以及19处错误修复。自v3.2.0以来，共修改117个文件，新增41,588行代码。\n\n### 支持的源类型（17种）\n\n| 序号 | 类型           | CLI命令         | 配置类型   | 自动检测       |\n|------|----------------|-----------------|------------|----------------|\n| 1    | 文档（网页）   | `scrape` \u002F `create \u003Curl>` | `documentation` | HTTP\u002FHTTPS URL |\n| 2    | GitHub仓库     | `github` \u002F `create owner\u002Frepo` | `github` | `owner\u002Frepo` 或 github.com URL |\n| 3    | PDF文档        | `pdf` \u002F `create file.pdf` | `pdf` | `.pdf` 扩展名 |\n| 4    | Word文档       | `word` \u002F `create file.docx` | `word` | `.docx` 扩展名 |\n| 5    | EPUB电子书     | `epub` \u002F `create file.epub` | `epub` | `.epub` 扩展名 |\n| 6    | 视频           | `video` \u002F `create \u003Curl\u002Ffile>` | `video` | YouTube\u002FVimeo URL、视频文件扩展名 |\n| 7    | 本地代码库     | `analyze` \u002F `create .\u002Fpath` | `local` | 目录路径 |\n| 8    | Jupyter Notebook | `jupyter` \u002F `create file.ipynb` | `jupyter` | `.ipynb` 扩展名 |\n| 9    | 本地HTML       | `html` \u002F `create file.html` | `html` | `.html`\u002F`.htm` 扩展名 |\n| 10   | OpenAPI\u002FSwagger  | `openapi` \u002F `create spec.yaml` | `openapi` | 包含OpenAPI内容的 `.yaml`\u002F`.yml` 文件 |\n| 11   | AsciiDoc       | `asciidoc` \u002F `create file.adoc` | `asciidoc` | `.adoc`\u002F`.asciidoc` 扩展名 |\n| 12   | PowerPoint     | `pptx` \u002F `create file.pptx` | `pptx` | `.pptx` 扩展名 |\n| 13   | RSS\u002FAtom订阅源 | `rss` \u002F `create feed.rss` | `rss` | `.rss`\u002F`.atom` 扩展名 |\n| 14   | Man手册页      | `manpage` \u002F `create cmd.1` | `manpage` | `.1`–`.8`\u002F`.man` 扩展名 |\n| 15   | Confluence维基 | `confluence` | `confluence` | API或导出目录 |\n| 16   | Notion页面     | `notion` | `notion` | API或导出目录 |\n| 17   | Slack\u002FDiscord聊天 | `chat` | `chat` | 导出目录或API |\n\n### 新增功能\n\n#### 10种新的Skill源类型（总计17种）\n\nSkill Seekers 现在支持17种源类型——相比之前的7种有了显著提升。每一种新类型都已全面集成到CLI中（`skill-seekers \u003Ctype>`）、`create`命令的自动检测机制、统一的多源配置、配置验证、MCP服务器以及技能构建器中。\n\n- **Jupyter Notebook** — `skill-seekers jupyter --notebook file.ipynb` 或 `skill-seekers create file.ipynb`\n  - 提取Markdown单元格、带输出的代码单元格、内核元数据、导入语句及语言检测。\n  - 支持单个文件和包含Notebook的目录；会过滤`.ipynb_checkpoints`文件。\n  - 可选依赖：`pip install \"skill-seekers[jupyter]\"`（nbformat）。\n  - 入口点：`skill-seekers-jupyter`。\n\n- **本地HTML** — `skill-seekers html --html-path file.html` 或 `skill-seekers create file.html`\n  - 使用BeautifulSoup解析HTML，并智能检测主要内容区域（如`\u003Carticle>`、`\u003Cmain>`、`.content`类或最大`div`）。\n  - 提取标题、代码块、表格（转换为Markdown格式）、图片、链接；将内联HTML转换为Markdown。\n  - 支持单个文件和目录；兼容`.html`、`.htm`、`.xhtml`等扩展名。\n  - 无需额外依赖。","2026-03-15T22:27:19",{"id":174,"version":175,"summary_zh":176,"released_at":177},62348,"v3.2.0","## v3.2.0 — 视频提取、Word支持、Pinecone适配器\n\n**主题：** 支持视频源、Word文档、Pinecone适配器，以及质量改进。自v3.1.3以来，共修改94个文件，新增23,500行代码。**2,540项测试通过。**\n\n### 🎬 视频提取流水线\n\n完整的视频提取系统，可将YouTube视频和本地视频文件转换为AI可消费的技能格式。\n\n- **`skill-seekers video --url \u003Cyoutube-url>`** — 新增用于视频抓取的CLI命令\n- **`skill-seekers create \u003Cyoutube-url>`** — 自动检测YouTube URL\n- **字幕提取** — 三层回退机制：YouTube API → yt-dlp → faster-whisper\n- **视觉OCR** — 多引擎集成（EasyOCR + pytesseract）用于代码帧\n- **面板检测** — 将IDE截图拆分为独立的子部分\n- **代码时间轴** — 通过编辑历史追踪代码在各帧中的演变\n- **两阶段AI增强** — 利用字幕上下文清理OCR噪声\n- **GPU自动检测** — `skill-seekers video --setup` 可检测CUDA\u002FROCm\u002FCPU，并安装正确的PyTorch版本\n- **197项测试** 覆盖模型、元数据、字幕、视觉、OCR及CLI\n\n### 📄 Word文档（.docx）支持\n\n- **`skill-seekers word --docx \u003Cfile>`** — 完整流水线：mammoth → HTML → 分节 → SKILL.md\n- **`skill-seekers create document.docx`** — 自动检测.docx文件\n- **智能代码检测** — 将等宽字体段落识别为代码块\n- **安装：** `pip install skill-seekers[docx]`\n\n### 🌲 Pinecone向量数据库适配器\n\n- **`skill-seekers package output\u002F --format pinecone --upload`** — 直接上传至Pinecone\n- 全面的CRUD操作，支持命名空间\n- 支持OpenAI和Sentence Transformers嵌入\n- 批量插入更新，可配置批次大小\n- **764项测试** 提供全面覆盖\n\n### 🐛 错误修复\n\n- **6项OCR质量修复** — 跳过摄像头帧、清理IDE装饰、修复重复行、过滤UI垃圾\n- **15项视频流水线修复** — 超时处理、MCP集成、文件名冲突、依赖管理\n- **问题#300** — 选择器回退与干运行链接发现（ReactFlow发现了20多页，原为1页）\n- **问题#301** — macOS下`setup.sh`修复\n- **RAG分块崩溃** — 修复了`AttributeError: output_dir`\n- **分块重叠自动缩放** — 缩放至`max(50, chunk_tokens \u002F\u002F 10)`\n- **移除引用文件限制** — 不再对GitHub议题、发布或代码块数量设上限\n- 更多详情请参阅[CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FCHANGELOG.md)\n\n### 📦 安装\u002F升级\n\n```bash\npip install --upgrade skill-seekers\n\n# 带视频支持\npip install skill-seekers[video]\nskill-seekers video --setup  # 自动检测GPU，安装依赖\n\n# 带Word支持\npip install skill-seekers[docx]\n\n# 带Pinecone支持\npip install skill-seekers[pinecone]\n\n# 全功能\npip install skill-seekers[all]\n```\n\n**完整变更日志：** https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FCHANGELOG.md","2026-03-02T09:44:00",{"id":179,"version":180,"summary_zh":181,"released_at":182},62349,"v3.1.3","## [3.1.3] - 2026-02-24\n\n### 🐛 热修复 — 显式分块标志及参数管道清理\n\n### 修复\n- **问题 #299: `skill-seekers package --target claude` 未识别参数导致崩溃** — `main.py` 中的 `_reconstruct_argv()` 函数在路由子命令时，会将默认标志值重新注入 argv。`package_skill.py` 内部有一个长达 105 行的内联参数解析器，其使用的标志名称与 `arguments\u002Fpackage.py` 中的不一致，因此转发的标志会被拒绝。现已通过用对 `add_package_arguments(parser)` 的调用替换该内联代码块来修复——使其成为唯一的事实来源。\n\n### 变更\n- **`package_skill.py` 参数解析器重构** — 将约 105 行的重复内联 argparse 代码替换为单一的 `add_package_arguments(parser)` 调用。标志名称现在可确保与 `_reconstruct_argv()` 的输出一致，从而避免未来出现参数名称不一致的问题。\n- **明确的分块标志名称** — 所有 `--chunk-*` 标志现均包含单位后缀，以消除 RAG 令牌与流式传输字符之间的歧义：\n  - `--chunk-size`（RAG 令牌）→ `--chunk-tokens`\n  - `--chunk-overlap`（RAG 令牌）→ `--chunk-overlap-tokens`\n  - `--chunk`（启用 RAG 分块）→ `--chunk-for-rag`\n  - `--streaming-chunk-size`（字符）→ `--streaming-chunk-chars`\n  - `--streaming-overlap`（字符）→ `--streaming-overlap-chars`\n  - PDF 提取器中的 `--chunk-size`（页数）→ `--pdf-pages-per-chunk`\n- **`setup_logging()` 中心化** — 在 `utils.py` 中新增了 `setup_logging(verbose, quiet)` 函数，并移除了 `doc_scraper.py`、`github_scraper.py`、`codebase_scraper.py` 和 `unified_scraper.py` 中四处重复的模块级 `logging.basicConfig()` 调用。\n\n","2026-02-24T19:57:09",{"id":184,"version":185,"summary_zh":186,"released_at":187},62350,"v3.1.2","## 变更内容\n\n### 🐛 重要错误修复\n\n**Gemini 增强功能 404 错误** — Google 已停用 `gemini-2.0-flash-exp` 模型，导致所有 Gemini 增强请求均返回 404 错误而失败。现已替换为 `gemini-2.5-flash`（稳定 GA 版本）。\n\n**`skill-seekers enhance` 自动检测功能** — 文档中说明的“当存在 API 密钥时自动使用 API 模式”的行为实际上并未实现。本次发布已修复此问题：\n- 设置了 `ANTHROPIC_API_KEY` → 使用 Claude API 模式\n- 设置了 `GOOGLE_API_KEY` → 使用 Gemini API 模式\n- 设置了 `OPENAI_API_KEY` → 使用 OpenAI API 模式\n- 未设置任何密钥 → 使用 LOCAL 模式（Claude Code Max，免费）\n\n即使设置了 API 密钥，仍可使用 `--mode LOCAL` 强制切换到本地模式。\n\n**`create` 命令参数传递问题** — 当与 GitHub、PDF 和代码库源一起使用时，通用标志（`--dry-run`、`--verbose`、`--quiet`、`--name`、`--description`）会导致程序崩溃。现已全部修复。此外，还为 `skill-seekers github` 和 `skill-seekers pdf` 添加了对 `--dry-run` 的支持。\n\n## 升级\n\n```bash\npip install --upgrade skill-seekers\n```\n\n```bash\ndocker pull yusufk\u002Fskill-seekers:latest\n```\n\n## 完整变更日志\n\n完整详情请参阅 [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FCHANGELOG.md)。","2026-02-24T04:09:57",{"id":189,"version":190,"summary_zh":191,"released_at":192},62351,"v3.1.1","## 变更内容\n* 修复：在 create 命令的 Web 路由中使用 getattr 获取 max_pages，由 @YusufKaraaslanSpyke 在 https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fpull\u002F294 中完成\n* 紧急修复：v3.1.1 — 修复 create 命令中的 max_pages AttributeError，由 @yusufkaraaslan 在 https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fpull\u002F295 中完成\n* 最大页数紧急修复，由 @yusufkaraaslan 在 https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fpull\u002F296 中完成\n\n\n**完整变更日志**：https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fcompare\u002Fv3.1.0...v3.1.1","2026-02-23T09:12:17",{"id":194,"version":195,"summary_zh":196,"released_at":197},62352,"v3.1.0","## 🎯 v3.1.0 — “统一的 CLI 与开发者体验”\n\n> **一个命令搞定一切。65 个工作流预设。178 个生产配置。2280+ 个测试。**\n\n---\n\n### 🚀 新功能\n\n#### ✨ 统一的 `create` 命令 — 一个命令通吃所有场景\n\n再也不用记着该用哪个命令了。只需使用 `create` 加上任何内容即可：\n\n```bash\n# 自动检测源类型\nskill-seekers create https:\u002F\u002Fdocs.react.dev\u002F          # → 网页爬虫\nskill-seekers create facebook\u002Freact                   # → GitHub 代码分析\nskill-seekers create .\u002Fmy-project                     # → 本地代码库\nskill-seekers create tutorial.pdf                     # → PDF 提取\nskill-seekers create configs\u002Freact.json               # → 多源统一处理\n\n# 快速预设快捷方式 (-p)\nskill-seekers create https:\u002F\u002Fdocs.react.dev\u002F -p quick\nskill-seekers create facebook\u002Freact -p comprehensive\n\n# 渐进式帮助 — 再也不被大量选项搞得不知所措\nskill-seekers create --help            # 13 个通用选项（简洁）\nskill-seekers create --help-web        # 针对 Web 的特定选项\nskill-seekers create --help-github     # 针对 GitHub 的特定选项\nskill-seekers create --help-all        # 所有选项（超过 120 个）\n```\n\n#### 🔧 65 个增强型工作流预设\n\n通过内置的工作流预设，为特定用例量身定制你的技能：\n\n```bash\n# 连接多个工作流\nskill-seekers create facebook\u002Freact \\\n  --enhance-workflow security-focus \\\n  --enhance-workflow api-documentation\n\n# 管理预设\nskill-seekers workflows list                    # 浏览所有 65 个内置预设\nskill-seekers workflows show security-focus     # 查看某个预设\nskill-seekers workflows copy security-focus     # 复制到用户目录以进行自定义\nskill-seekers workflows add my-preset.yaml      # 添加自定义预设\n```\n\n内置预设涵盖：`security-focus`、`api-documentation`、`architecture-comprehensive`、`testing-focus`、`microservices-patterns`、`kubernetes-deployment`、`database-schema`、`mlops-pipeline`、`rest-api-design`、`graphql-schema`、`responsive-design`、`performance-optimization`、`accessibility-a11y`，以及 [50 多种其他预设](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Ftree\u002Fmain\u002Fsrc\u002Fskill_seekers\u002Fworkflows)。\n\n#### ⚡ 智能增强调度器\n\n```bash\n# 自动检测 API 密钥，或回退到 Claude Code CLI\nskill-seekers enhance output\u002Freact\u002F\n\n# 显式指定目标\nskill-seekers enhance output\u002Freact\u002F --target gemini\n\n# Docker\u002Froot 保护机制 — 清晰报错而非静默失败\n# （修复 #286、#289）\n```\n\n#### 📄 支持 ReStructuredText (RST)\n\nSphinx\u002FRST 文档站点现在可以正确提取内容——类引用、代码块、表格和交叉引用都能被正确解析。\n\n---\n\n### 🗃️ 178 个生产配置 — 全部经过审核与优化\n\n[skill-seekers-configs](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002Fskill-seekers-configs) 中的所有配置均已达到 **v1.1.0 质量标准**：\n\n- ✅ 移除了所有 `max_pages` 字段（已弃用，将自动应用默认值）\n- ✅ 每个配置包含 5–13 个类别，每个类别 3–6 个关键词\n- ✅ 语义选择器后备链（articl","2026-02-22T22:37:03",{"id":199,"version":200,"summary_zh":201,"released_at":202},62353,"v3.0.0","## [3.0.0] - 2026-02-10\n\n### 🚀 “通用智能平台”——重大发布\n\n**主题：** 将任何文档转化为适用于任何AI系统的结构化知识。\n\n这是我们迄今为止最大的一次发布！v3.0.0版本确立了Skill Seekers作为整个AI生态系统的**通用文档预处理器**——从RAG流水线到AI编码助手，再到Claude技能。\n\n### 亮点\n\n- 🚀 **16个平台适配器**（相比v2.x版本的4个有所增加）\n- 🛠️ **26种MCP工具**（相比之前的9种有所增加）\n- ✅ **1,852项测试通过**（相比之前的700多项有所提升）\n- ☁️ **云存储**支持（S3、GCS、Azure）\n- 🔄 **CI\u002FCD就绪**（GitHub Action + Docker）\n- 📦 **12个示例项目**，覆盖每种集成场景\n- 📚 **18份集成指南**已完成\n\n### 新增——平台适配器（共16个）\n\n#### RAG与向量数据库（8个）\n- **LangChain** (`--format langchain`) - 输出LangChain Document对象\n- **LlamaIndex** (`--format llama-index`) - 输出LlamaIndex TextNode对象\n- **Chroma** (`--format chroma`) - 直接集成ChromaDB\n- **FAISS** (`--format faiss`) - Facebook AI相似性搜索\n- **Haystack** (`--format haystack`) - Deepset Haystack流水线\n- **Qdrant** (`--format qdrant`) - Qdrant向量数据库\n- **Weaviate** (`--format weaviate`) - Weaviate向量搜索\n- **Pinecone就绪** (`--target markdown`) - Markdown格式已为Pinecone做好准备\n\n#### AI平台（3个）\n- **Claude** (`--target claude`) - Claude AI技能（ZIP + YAML）\n- **Gemini** (`--target gemini`) - Google Gemini技能（tar.gz）\n- **OpenAI** (`--target openai`) - OpenAI ChatGPT（ZIP + 向量存储）\n\n#### AI编码助手（4个）\n- **Cursor** (`--target claude` + `.cursorrules`) - Cursor IDE集成\n- **Windsurf** (`--target claude` + `.windsurfrules`) - Windsurf\u002FCodeium\n- **Cline** (`--target claude` + `.clinerules`) - VS Code扩展\n- **Continue.dev** (`--target claude`) - 通用IDE支持\n\n#### 通用型（1个）\n- **Markdown** (`--target markdown`) - 通用ZIP导出格式\n\n### 新增——MCP工具（共26个）\n\n#### 配置工具（3个）\n- `generate_config` - 生成爬取配置\n- `list_configs` - 列出可用的预设配置\n- `validate_config` - 验证配置JSON结构\n\n#### 爬取工具（8个）\n- `estimate_pages` - 在爬取前估算页面数量\n- `scrape_docs` - 爬取文档网站\n- `scrape_github` - 爬取GitHub仓库\n- `scrape_pdf` - 从PDF文件中提取内容\n- `scrape_codebase` - 分析本地代码库\n- `detect_patterns` - 检测代码中的设计模式\n- `extract_test_examples` - 从测试中提取使用示例\n- `build_how_to_guides` - 根据代码构建操作指南\n\n#### 打包工具（4个）\n- `package_skill` - 为目标平台打包技能\n- `upload_skill` - 上传至LLM平台\n- `enhance_skill` - 基于AI的增强功能\n- `install_skill` - 一键式完整工作流程\n\n#### 源工具（5个）\n- `fetch_config` - 从远程源获取配置\n- `submit_config` - 提交配置以待审批\n- `add_config_source` - 添加Git配置源\n- `list_config_sources` - 列出配置源","2026-02-08T20:30:33",{"id":204,"version":205,"summary_zh":206,"released_at":207},62354,"v2.9.0","## 🎮 游戏开发发布 - Godot 引擎支持\n\n本次发布新增对 **Godot 游戏引擎** 项目的全面支持，提供业界领先的信号流分析功能，并完整支持 GDScript 语言。\n\n### 🎮 新增功能\n\n#### C3.10：Godot 项目信号流分析 ⭐ 新增\n- **完整的信号流分析系统**：分析 Godot 游戏项目中的事件驱动架构\n  - 信号声明提取（`signal` 关键字检测）\n  - 连接映射（`.connect()` 调用及其目标和方法）\n  - 发射追踪（`.emit()` 和 `emit_signal()` 调用）\n  - **实际结果**：在 Cosmic Idler 测试项目中检测到 208 个信号、634 条连接和 298 次发射\n  - 信号密度指标（0.78 个信号\u002F文件）\n  - 事件链检测（信号触发其他信号）\n  - 输出：`signal_flow.json`（374KB）、`signal_flow.mmd`（Mermaid 图）、`signal_reference.md`（34KB）\n\n- **信号模式检测**：以置信度评分识别出三种主要模式\n  - **EventBus 模式**（置信度 0.90）：自动加载脚本中的集中式信号中心\n  - **观察者模式**（置信度 0.85）：多观察者信号（3 个以上监听器，如 `theme_changed` 共有 21 条连接）\n  - **事件链**（置信度 0.80）：级联式的信号传播\n\n- **基于信号的使用指南（C3.10.1）**：由 AI 生成的使用指南\n  - 分步指南（连接 → 发射 → 处理）\n  - 来自项目的实际代码示例\n  - 常见使用位置及文件引用\n  - 参数文档\n  - 输出：`signal_how_to_guides.md`（为 Cosmic Idler 生成了 10 篇指南）\n\n#### 完整的 Godot 游戏引擎支持\n- **全面的 Godot 文件类型支持**：完整分析 Godot 4.x 项目\n  - **GDScript (.gd)**：测试项目中共分析 265 个文件（占代码库的 59.8%）\n  - **场景文件 (.tscn)**：118 个场景文件（占 26.6%）\n  - **资源文件 (.tres)**：38 个资源文件（占 8.6%）\n  - **着色器文件 (.gdshader, .gdshaderinc)**：9 个着色器文件（占 2.0%）\n  - **C# 集成**：Phantom Camera 插件（13 个文件，占 2.9%）\n\n- **GDScript 语言支持**：采用正则表达式进行完整解析与提取\n  - 依赖项提取：`preload()`、`load()`、`extends` 等模式\n  - 测试框架检测：GUT、gdUnit4、WAT\n  - 测试文件模式：`test_*.gd`、`*_test.gd`\n  - 信号语法：`signal`、`.connect()`、`.emit()`\n  - 导出装饰器：`@export`、`@onready`\n  - 测试装饰器：`@test`（gdUnit4）\n  - 共提取 377 个依赖项，未出现任何语法错误。\n\n- **游戏引擎框架检测**：改进了对 Unity、Unreal 和 Godot 的检测能力\n  - **Godot 标识**：`project.godot`、`.godot` 目录、`.tscn`、`.tres`、`.gd` 文件\n  - **Unity 标识**：`Assembly-CSharp.csproj`、`UnityEngine.dll`、`ProjectSettings\u002FProjectVersion.txt`\n  - **Unreal 标识**：`.uproject`、`Source\u002F`、`Config\u002FDefaultEngine.ini`\n  - 修复了误判 Unity 的问题（此前误用通用关键词“Assets”）\n  - 优先级检测机制（游戏引擎优先于 Web 框架检测）。\n\n- **GDScript 测试用例提取**：提取用例","2026-02-02T20:35:45",{"id":209,"version":210,"summary_zh":211,"released_at":212},62355,"v2.8.0","## [2.8.0] - 2026-02-01\n\n### 🚀 重大功能发布 - 增强的代码分析与文档生成\n\n本次发布带来了强大的新代码分析功能、性能优化以及国际化 API 支持。特别感谢所有为此次发布贡献力量的贡献者！\n\n### 新增功能\n\n#### C3.9：项目文档提取\n- **Markdown 文档提取**：自动从项目中提取并分类所有 `.md` 文件\n  - 根据文件夹\u002F文件名进行智能分类（概览、架构、指南、工作流、特性等）\n  - 提供处理深度控制：`surface`（原始复制）、`deep`（解析+摘要）、`full`（AI增强）\n  - AI 增强（等级 2 及以上）会添加主题提取和交叉引用\n  - SKILL.md 中新增“📖 项目文档”章节\n  - 输出至 `references\u002Fdocumentation\u002F` 目录，并按类别组织\n  - 默认开启，使用 `--skip-docs` 可禁用\n  - 新增 15 个用于文档提取功能的测试用例\n\n#### 细粒度 AI 增强控制\n- **`--enhance-level` 标志**：对 AI 增强进行细粒度控制（0-3）\n  - 等级 0：无 AI 增强（默认）\n  - 等级 1：仅增强 SKILL.md（快速且高价值）\n  - 等级 2：SKILL.md + 架构 + 配置 + 文档\n  - 等级 3：全面增强（模式、测试、配置、架构、文档）\n- **配置集成**：在 `~\u002F.config\u002Fskill-seekers\u002Fconfig.json` 中新增 `default_enhance_level` 设置\n- **MCP 支持**：所有 MCP 工具均已更新，支持 `enhance_level` 参数\n- **与 `--comprehensive` 独立**：增强级别与功能深度相互独立\n\n#### C# 语言支持\n- **C# 测试示例提取**：全面支持 C# 测试框架\n  - 语言别名映射（C# → csharp，C++ → cpp）\n  - NUnit、xUnit、MSTest 测试框架模式\n  - 模拟模式支持（NSubstitute、Moq）\n  - Zenject 依赖注入模式\n  - 设置\u002F清理方法提取\n  - 新增 2 个用于 C# 提取功能的测试用例\n\n#### 性能优化\n- **并行 LOCAL 模式 AI 增强**：使用 ThreadPoolExecutor 后速度提升 6-12 倍\n  - 并发工作线程数：3 个（可通过 `local_parallel_workers` 配置）\n  - 批量处理：每次 Claude CLI 调用处理 20 个模式（可通过 `local_batch_size` 配置）\n  - 对大型代码库有显著加速效果\n- **配置设置**：在配置中新增 `ai_enhancement` 部分\n  - `local_batch_size`：每次 CLI 调用处理的模式数量（默认：20）\n  - `local_parallel_workers`：并发工作线程数（默认：3）\n\n#### 用户体验改进\n- **自动增强**：当使用 `--enhance` 或 `--comprehensive` 时，SKILL.md 会自动增强\n  - 无需再单独运行 `skill-seekers enhance` 命令\n  - 实现无缝的一键工作流程\n  - 大型代码库超时时间为 10 分钟\n  - 出现失败时会优雅降级，并提供重试说明\n- **LOCAL 模式回退**：所有 AI 增强功能在未设置 API 密钥时都会回退到 LOCAL 模式\n  - 适用于：模式增强（C3.1）、测试示例（C3.2）、架构（C3.7）\n  - 使用 Claude Code CLI 替代静默失败\n  - 更好的用户体验：“正在使用 LOCAL 模式","2026-02-02T20:34:15",{"id":214,"version":215,"summary_zh":216,"released_at":217},62356,"v2.7.4","## 🔧 Bug Fix - Language Selector Links\n\nThis **patch release** fixes the broken Chinese language selector link that appeared on PyPI and other non-GitHub platforms.\n\n### Fixed\n\n- **Broken Language Selector Links on PyPI**\n  - **Issue**: Chinese language link used relative URL (`README.zh-CN.md`) which only worked on GitHub\n  - **Impact**: Users on PyPI clicking \"简体中文\" got 404 errors\n  - **Solution**: Changed to absolute GitHub URL\n  - **Result**: Language selector now works on PyPI, GitHub, and all platforms\n  - **Files Fixed**: `README.md`, `README.zh-CN.md`\n\n### Links\n\n- **PyPI Package**: https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F2.7.4\u002F\n- **Full Changelog**: https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FCHANGELOG.md#274---2026-01-22","2026-01-21T21:14:28",{"id":219,"version":220,"summary_zh":221,"released_at":222},62357,"v2.7.3","# 🌏 International i18n Release\n\nThis **documentation release** adds comprehensive Chinese language support, making Skill Seekers accessible to the world's largest developer community.\n\n## ✨ What's New\n\n### 🇨🇳 Chinese (Simplified) Documentation\n- **Complete README Translation** - 1,962 lines of comprehensive Chinese documentation (README.zh-CN.md)\n- **Language Selector Badges** - Easy switching between English and Chinese in both READMEs\n- **Machine Translation Disclaimer** - Honest labeling with invitation for community improvements\n- **Community Engagement** - GitHub issue #260 created for native speakers to improve translation quality\n\n### 📦 PyPI Metadata Internationalization\n- **Updated Package Description** - Now highlights Chinese documentation availability\n- **i18n Keywords** - Added \"i18n\", \"chinese\", \"international\" for better discoverability\n- **Natural Language Classifiers** - English and Chinese (Simplified) officially declared\n- **Direct Chinese README Link** - Added to project URLs for easy access from PyPI\n\n## 🌍 Why This Matters\n\n**Market Impact:**\n- ✅ Reaches 1+ billion Chinese speakers worldwide\n- ✅ Taps into the world's largest developer community\n- ✅ Better discoverability on Chinese search engines (Baidu, Gitee, etc.)\n- ✅ Professional image showing international awareness\n- ✅ Competitive advantage - most similar tools lack Chinese documentation\n\n**For Users:**\n- ✅ Native language documentation lowers barrier to entry\n- ✅ Better user experience with familiar terminology\n- ✅ Increased engagement from Chinese developer community\n- ✅ Potential for more contributors and feedback\n\n## 🤝 Community Contribution\n\nWe invite Chinese developers to help improve the translation:\n\n- **Review Issue**: [#260](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fissues\u002F260)\n- **What to Review**: Technical accuracy, natural expression, terminology\n- **How to Help**: Comment on the issue with suggestions or submit a PR\n\nAll contributions are welcome and appreciated!\n\n## 📥 Installation\n\n\n\n## 🔗 Important Links\n\n- **Chinese README**: [README.zh-CN.md](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FREADME.zh-CN.md)\n- **Community Review**: [Issue #260](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fissues\u002F260)\n- **PyPI Package**: https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F2.7.3\u002F\n- **Official Website**: https:\u002F\u002Fskillseekersweb.com\u002F\n\n## 📝 Full Changelog\n\nSee [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FCHANGELOG.md#273---2026-01-21) for complete release notes.\n\n---\n\n**语言 \u002F Languages:**\n- [English](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FREADME.md)\n- [简体中文](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FREADME.zh-CN.md)","2026-01-21T21:02:26",{"id":224,"version":225,"summary_zh":226,"released_at":227},62358,"v2.7.2","## 🚨 Critical CLI Bug Fixes\n\nThis **hotfix release** resolves 4 critical CLI bugs reported in issues #258 and #259 that prevented core commands from working correctly.\n\n### Fixed\n\n**Issue #258: `install --config` command fails with unified scraper** (#258)\n- **Root Cause**: `unified_scraper.py` missing `--fresh` and `--dry-run` argument definitions\n- **Solution**: Added both flags to unified_scraper argument parser and main.py dispatcher\n- **Impact**: `skill-seekers install --config react` now works without \"unrecognized arguments\" error\n\n**Issue #259 (Original): `scrape` command doesn't accept URL and --max-pages** (#259)\n- **Root Cause**: No positional URL argument or `--max-pages` flag support\n- **Solution**: Added positional URL argument and `--max-pages` flag with safety warnings\n- **Impact**: `skill-seekers scrape https:\u002F\u002Fexample.com --max-pages 50` now works\n- **Safety Warnings**: Warns if max-pages > 1000 or \u003C 10\n\n**Issue #259 (Comment A): Version shows 2.7.0 instead of actual version** (#259)\n- **Root Cause**: Hardcoded version string in main.py\n- **Solution**: Import `__version__` from `__init__.py` dynamically\n- **Impact**: `skill-seekers --version` now shows correct version (2.7.2)\n\n**Issue #259 (Comment B): PDF command shows empty \"Error: \" message** (#259)\n- **Root Cause**: Exception handler didn't handle empty exception messages\n- **Solution**: Improved exception handler to show exception type and added context-specific messages\n- **Impact**: PDF errors now show clear messages instead of just \"Error: \"\n\n### Installation\n\n```bash\npip install --upgrade skill-seekers\n```\n\n### Testing\n\n- ✅ Verified all commands work with exact issue reproduction steps\n- ✅ All 202 tests passing\n\n### Full Changelog\nhttps:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FCHANGELOG.md#272---2026-01-21","2026-01-21T20:24:16",{"id":229,"version":230,"summary_zh":231,"released_at":232},62359,"v2.7.1","## 🚨 Critical Bug Fix - Config Download 404 Errors\n\nThis **hotfix release** resolves a critical bug causing 404 errors when downloading configs from the API.\n\n### Fixed\n\n- **Critical: Config download 404 errors** - Fixed bug where code was constructing download URLs manually instead of using the `download_url` field from the API response\n  - **Root Cause**: Code was building `f\"{API_BASE_URL}\u002Fapi\u002Fdownload\u002F{config_name}.json\"` which failed when actual URLs differed (CDN URLs, version-specific paths)\n  - **Solution**: Changed to use `config_info.get(\"download_url\")` from API response in both MCP server implementations\n  - **Files Fixed**:\n    - `src\u002Fskill_seekers\u002Fmcp\u002Ftools\u002Fsource_tools.py` (FastMCP server)\n    - `src\u002Fskill_seekers\u002Fmcp\u002Fserver_legacy.py` (Legacy server)\n  - **Impact**: Fixes all config downloads from skillseekersweb.com API and private Git repositories\n  - **Reported By**: User testing `skill-seekers install --config godot --unlimited`\n  - **Testing**: All 15 source tools tests pass, all 8 fetch_config tests pass\n\n### Installation\n\n```bash\npip install --upgrade skill-seekers\n```\n\nOr install a specific version:\n\n```bash\npip install skill-seekers==2.7.1\n```\n\n### Links\n\n- **PyPI**: https:\u002F\u002Fpypi.org\u002Fproject\u002Fskill-seekers\u002F2.7.1\u002F\n- **Website**: https:\u002F\u002Fskillseekersweb.com\u002F\n- **Documentation**: https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FREADME.md\n\n---\n\n🤖 Generated with [Claude Code](https:\u002F\u002Fclaude.com\u002Fclaude-code)\n\nCo-Authored-By: Claude Sonnet 4.5 \u003Cnoreply@anthropic.com>","2026-01-18T19:40:56",{"id":234,"version":235,"summary_zh":236,"released_at":237},62360,"v2.7.0","## [2.7.0] - 2026-01-18\n\n### 🔐 Smart Rate Limit Management & Multi-Token Configuration\n\nThis **minor feature release** introduces intelligent GitHub rate limit handling, multi-profile token management, and comprehensive configuration system. Say goodbye to indefinite waits and confusing token setup!\n\n### Added\n\n- **🎯 Multi-Token Configuration System** - Flexible GitHub token management with profiles\n  - **Secure config storage** at `~\u002F.config\u002Fskill-seekers\u002Fconfig.json` with 600 permissions\n  - **Multiple GitHub profiles** support (personal, work, OSS, etc.)\n    - Per-profile rate limit strategies: `prompt`, `wait`, `switch`, `fail`\n    - Configurable timeout per profile (default: 30 minutes)\n    - Auto-detection and smart fallback chain\n    - Profile switching when rate limited\n  - **API key management** for Claude, Gemini, OpenAI\n    - Environment variable fallback (ANTHROPIC_API_KEY, GOOGLE_API_KEY, OPENAI_API_KEY)\n    - Config file storage with secure permissions\n  - **Progress tracking** for resumable jobs\n    - Auto-save at configurable intervals (default: 60 seconds)\n    - Job metadata: command, progress, checkpoints, timestamps\n    - Stored at `~\u002F.local\u002Fshare\u002Fskill-seekers\u002Fprogress\u002F`\n  - **Auto-cleanup** of old progress files (default: 7 days, configurable)\n  - **First-run experience** with welcome message and quick setup\n  - **ConfigManager class** with singleton pattern for global access\n\n- **🧙 Interactive Configuration Wizard** - Beautiful terminal UI for easy setup\n  - **Main menu** with 7 options:\n    1. GitHub Token Setup\n    2. API Keys (Claude, Gemini, OpenAI)\n    3. Rate Limit Settings\n    4. Resume Settings\n    5. View Current Configuration\n    6. Test Connections\n    7. Clean Up Old Progress Files\n  - **GitHub token management**:\n    - Add\u002Fremove profiles with descriptions\n    - Set default profile\n    - Browser integration - opens GitHub token creation page\n    - Token validation with format checking (ghp_*, github_pat_*)\n    - Strategy selection per profile\n  - **API keys setup** with browser integration for each provider\n  - **Connection testing** to verify tokens and API keys\n  - **Configuration display** with current status and sources\n  - **CLI commands**:\n    - `skill-seekers config` - Main menu\n    - `skill-seekers config --github` - Direct to GitHub setup\n    - `skill-seekers config --api-keys` - Direct to API keys\n    - `skill-seekers config --show` - Show current config\n    - `skill-seekers config --test` - Test connections\n\n- **🚦 Smart Rate Limit Handler** - Intelligent GitHub API rate limit management\n  - **Upfront warning** about token status (60\u002Fhour vs 5000\u002Fhour)\n  - **Real-time detection** of rate limits from GitHub API responses\n    - Parses X-RateLimit-* headers\n    - Detects 403 rate limit errors\n    - Calculates reset time from timestamps\n  - **Live countdown timers** with progress display\n  - **Automatic profile switching** - tries next available profile when rate limited\n  - **Four rate limit strategies**:\n    - `prompt` - Ask user what to do (default, interactive)\n    - `wait` - Auto-wait with countdown timer\n    - `switch` - Automatically try another profile\n    - `fail` - Fail immediately with clear error\n  - **Non-interactive mode** for CI\u002FCD (fail fast, no prompts)\n  - **Configurable timeouts** per profile (prevents indefinite waits)\n  - **RateLimitHandler class** with strategy pattern\n  - **Integration points**: GitHub fetcher, GitHub scraper\n\n- **📦 Resume Command** - Resume interrupted scraping jobs\n  - **List resumable jobs** with progress details:\n    - Job ID, started time, command\n    - Current phase and file counts\n    - Last updated timestamp\n  - **Resume from checkpoints** (skeleton implemented, ready for integration)\n  - **Auto-cleanup** of old jobs (respects config settings)\n  - **CLI commands**:\n    - `skill-seekers resume --list` - List all resumable jobs\n    - `skill-seekers resume \u003Cjob-id>` - Resume specific job\n    - `skill-seekers resume --clean` - Clean up old jobs\n  - **Progress storage** at `~\u002F.local\u002Fshare\u002Fskill-seekers\u002Fprogress\u002F\u003Cjob-id>.json`\n\n- **⚙️ CLI Enhancements** - New flags and improved UX\n  - **--non-interactive flag** for CI\u002FCD mode\n    - Available on: `skill-seekers github`\n    - Fails fast on rate limits instead of prompting\n    - Perfect for automated pipelines\n  - **--profile flag** to select specific GitHub profile\n    - Available on: `skill-seekers github`\n    - Uses configured profile from `~\u002F.config\u002Fskill-seekers\u002Fconfig.json`\n    - Overrides environment variables and defaults\n  - **Entry points** for new commands:\n    - `skill-seekers-config` - Direct config command access\n    - `skill-seekers-resume` - Direct resume command access\n\n- **🧪 Comprehensive Test Suite** - Full test coverage for new features\n  - **16 new tests** in `test_rate_limit_handler.py`\n  - **Test coverage**:\n    - Header creation (with\u002Fwithout token)\n    - Handler initialization (token, strategy, config)\n    - Rate limit detection and extraction\n    - Upfr","2026-01-18T11:32:23",{"id":239,"version":240,"summary_zh":241,"released_at":242},62361,"v2.6.0","## 🚀 Complete C3.x Codebase Analysis Suite + Documentation Reorganization\n\nThis is a **major feature release** that delivers the complete C3.x codebase analysis suite (C3.1-C3.8), transforming Skill Seekers into a comprehensive code documentation and analysis tool. Also includes comprehensive documentation reorganization and quality-of-life improvements.\n\n---\n\n## 🎯 Complete C3.x Codebase Analysis Suite\n\n### C3.1 Design Pattern Detection\n- **10 Design Patterns**: Singleton, Factory, Observer, Strategy, Decorator, Builder, Adapter, Command, Template Method, Chain of Responsibility\n- **9 Languages**: Python, JavaScript, TypeScript, C++, C, C#, Go, Rust, Java (plus Ruby, PHP)\n- **3 Detection Levels**: Surface (fast), deep (balanced), full (thorough)\n- **CLI**: `skill-seekers-patterns --file src\u002Fdb.py`\n- **87% precision, 80% recall** (tested on 100 real-world projects)\n\n### C3.2 Test Example Extraction\n- **Extracts real usage examples** from test files\n- **5 Categories**: instantiation, method_call, config, setup, workflow\n- **9 Languages**: Python (AST-based), JavaScript, TypeScript, Go, Rust, Java, C#, PHP, Ruby\n- **Quality filtering** with confidence scoring\n- **CLI**: `skill-seekers extract-test-examples tests\u002F --language python`\n\n### C3.3 How-To Guide Generation with AI Enhancement ⭐\n- **Transforms test workflows** into step-by-step educational guides\n- **🆕 COMPREHENSIVE AI ENHANCEMENT** - 5 automatic improvements:\n  1. Step Descriptions - Natural language explanations\n  2. Troubleshooting Solutions - Diagnostic flows + solutions\n  3. Prerequisites Explanations - Why needed + setup instructions\n  4. Next Steps Suggestions - Related guides, learning paths\n  5. Use Case Examples - Real-world scenarios\n- **3 AI Modes**:\n  - **API Mode**: Claude API (requires ANTHROPIC_API_KEY)\n  - **LOCAL Mode**: Claude Code CLI (FREE, no API key needed!)\n  - **AUTO Mode**: Automatic detection (default)\n- **Quality Transformation**: 75-line templates → 500+ line professional tutorials\n- **CLI**: `skill-seekers-how-to-guides test_examples.json --ai-mode auto`\n\n### C3.4 Configuration Pattern Extraction with AI Enhancement\n- **9 Config Formats**: JSON, YAML, TOML, ENV, INI, Python, JS\u002FTS, Dockerfile, Docker Compose\n- **7 Common Patterns**: Database, API, Logging, Cache, Email, Auth, Server configs\n- **🆕 AI ENHANCEMENT** (optional):\n  1. Explanations - What each setting does\n  2. Best Practices - Suggested improvements\n  3. Security Analysis - Identifies hardcoded secrets\n  4. Migration Suggestions - Consolidation opportunities\n  5. Context - Pattern explanations\n- **CLI**: `skill-seekers-config-extractor --directory . --enhance-local`\n\n### C3.5 Architectural Overview & Skill Integrator\n- **ARCHITECTURE.md Generation** - Comprehensive architectural overview with 8 sections:\n  1. Overview, 2. Architectural Patterns, 3. Technology Stack, 4. Design Patterns\n  5. Configuration Overview, 6. Common Workflows, 7. Usage Examples, 8. Entry Points\n- **Default ON** - Runs automatically when GitHub sources have `local_repo_path`\n- **Organized outputs** in `references\u002Fcodebase_analysis\u002F`\n- **Enhanced SKILL.md** with Architecture & Code Analysis summary\n\n### C3.6 AI Enhancement\n- **AI-powered insights** for patterns and test examples\n- **Pattern Enhancement**: Explains why patterns detected, suggests improvements\n- **Test Example Enhancement**: Adds context, groups into tutorials, identifies best practices\n- **Batch processing** (5 items per call) for efficiency\n\n### C3.7 Architectural Pattern Detection\n- **Detects high-level patterns**: MVC, MVVM, MVP, Repository, Service Layer, Layered, Clean Architecture\n- **Framework detection**: Django, Flask, Spring, ASP.NET, Rails, Laravel, Angular, React, Vue.js\n- **Evidence-based** with confidence scoring\n- **AI-enhanced** architectural recommendations\n\n### C3.8 Standalone Codebase Scraper SKILL.md Generation\n- **Generates comprehensive SKILL.md** (300+ lines) with all C3.x analysis integrated\n- **Sections**: Description, When to Use, Quick Reference, Design Patterns, Architecture, Configuration\n- **Perfect for**: Private codebases, offline analysis, local project documentation\n- **CLI**: `skill-seekers-codebase-scraper --directory \u002Fpath\u002Fto\u002Fcode`\n\n---\n\n## ✨ Enhanced LOCAL Enhancement Modes\n\n**4 Execution Modes** for different use cases:\n- **Headless** (default): Foreground, waits for completion (perfect for CI\u002FCD)\n- **Background** (`--background`): Background thread, returns immediately\n- **Daemon** (`--daemon`): Fully detached with `nohup`, survives parent exit\n- **Terminal** (`--interactive-enhancement`): Opens new terminal window (macOS)\n\n**Force Mode (Default ON)**: Skip all confirmations - perfect for CI\u002FCD automation!\n\n**Status Monitoring**: New `enhance-status` command for background\u002Fdaemon processes\n- `skill-seekers enhance-status output\u002Freact\u002F` - Check status\n- `skill-seekers enhance-status output\u002Freact\u002F --watch` - Real-time watch\n- `skill-seekers enhance-status output\u002Freact\u002F --json` - JSON ou","2026-01-13T20:15:45",{"id":244,"version":245,"summary_zh":246,"released_at":247},62362,"v2.5.1","## [2.5.1] - 2025-12-30\n\n### 🐛 Critical Bug Fix - PyPI Package Broken\n\nThis **patch release** fixes a critical packaging bug that made v2.5.0 completely unusable for PyPI users.\n\n### Fixed\n\n- **CRITICAL**: Added missing `skill_seekers.cli.adaptors` module to packages list in pyproject.toml ([#221](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fpull\u002F221))\n  - **Issue**: v2.5.0 on PyPI throws `ModuleNotFoundError: No module named 'skill_seekers.cli.adaptors'`\n  - **Impact**: Broke 100% of multi-platform features (Claude, Gemini, OpenAI, Markdown)\n  - **Cause**: The adaptors module was missing from the explicit packages list\n  - **Fix**: Added `skill_seekers.cli.adaptors` to packages in pyproject.toml\n  - **Credit**: Thanks to [@MiaoDX](https:\u002F\u002Fgithub.com\u002FMiaoDX) for finding and fixing this issue!\n\n### Package Structure\n\nThe `skill_seekers.cli.adaptors` module contains the platform adaptor architecture:\n- `base.py` - Abstract base class for all adaptors\n- `claude.py` - Claude AI platform implementation\n- `gemini.py` - Google Gemini platform implementation\n- `openai.py` - OpenAI ChatGPT platform implementation\n- `markdown.py` - Generic markdown export\n\n**Note**: v2.5.0 is broken on PyPI. All users should upgrade to v2.5.1 immediately.\n\n---\n\n","2025-12-30T20:36:15",{"id":249,"version":250,"summary_zh":251,"released_at":252},62363,"v2.5.0","# 🚀 Multi-Platform Feature Parity - 4 LLM Platforms Supported\n\nThis **major feature release** adds complete multi-platform support for **Claude AI**, **Google Gemini**, **OpenAI ChatGPT**, and **Generic Markdown** export. All features now work across all platforms with full feature parity.\n\n## 🎯 Highlights\n\n### Multi-LLM Platform Support\n- ✅ **4 platforms supported**: Claude AI, Google Gemini, OpenAI ChatGPT, Generic Markdown\n- ✅ **Complete feature parity**: All skill modes work with all platforms\n- ✅ **Platform adaptors**: Clean architecture with platform-specific implementations\n- ✅ **Unified workflow**: Same scraping output works for all platforms\n- ✅ **Smart enhancement**: Platform-specific AI models (Claude Sonnet 4, Gemini 2.0 Flash, GPT-4o)\n\n### Platform-Specific Capabilities\n\n| Platform | Format | Upload | Enhancement | Unique Features |\n|----------|--------|--------|-------------|----------------|\n| **Claude AI** | ZIP + YAML | Skills API | Sonnet 4 | MCP integration |\n| **Google Gemini** | tar.gz | Files API | Gemini 2.0 | 1M token context |\n| **OpenAI ChatGPT** | ZIP + Vector | Assistants API | GPT-4o | Semantic search |\n| **Generic Markdown** | ZIP | Manual | - | Universal compatibility |\n\n### Complete Feature Parity\n\n**All skill modes work with all platforms:**\n- 📄 Documentation scraping → All 4 platforms\n- 🐙 GitHub repository analysis → All 4 platforms\n- 📕 PDF extraction → All 4 platforms\n- 🔀 Unified multi-source → All 4 platforms\n- 💻 Local repository analysis → All 4 platforms\n\n### 18 MCP Tools with Multi-Platform Support\n- `package_skill` - Now accepts `target` parameter (claude, gemini, openai, markdown)\n- `upload_skill` - Now accepts `target` parameter (claude, gemini, openai)\n- `enhance_skill` - **NEW** standalone tool with `target` parameter\n- `install_skill` - Full multi-platform workflow automation\n\n## 📦 Installation\n\n```bash\n# Core package (Claude support)\npip install skill-seekers==2.5.0\n\n# With Gemini support\npip install skill-seekers[gemini]==2.5.0\n\n# With OpenAI support\npip install skill-seekers[openai]==2.5.0\n\n# With all platforms\npip install skill-seekers[all-llms]==2.5.0\n```\n\n## 🚀 Quick Start - Multi-Platform\n\n```bash\n# Scrape documentation (platform-agnostic)\nskill-seekers scrape --config configs\u002Freact.json\n\n# Package for different platforms\nskill-seekers package output\u002Freact\u002F --target claude     # ZIP\nskill-seekers package output\u002Freact\u002F --target gemini     # tar.gz\nskill-seekers package output\u002Freact\u002F --target openai     # ZIP with vector\nskill-seekers package output\u002Freact\u002F --target markdown   # ZIP universal\n\n# Upload to platforms (requires API keys)\nexport ANTHROPIC_API_KEY=sk-ant-...\nexport GOOGLE_API_KEY=AIzaSy...\nexport OPENAI_API_KEY=sk-proj-...\n\nskill-seekers upload output\u002Freact.zip --target claude\nskill-seekers upload output\u002Freact-gemini.tar.gz --target gemini\nskill-seekers upload output\u002Freact-openai.zip --target openai\n```\n\n## 📚 Documentation\n\n- 📊 [Complete Feature Matrix](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002Fdocs\u002FFEATURE_MATRIX.md)\n- 📤 [Multi-Platform Upload Guide](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002Fdocs\u002FUPLOAD_GUIDE.md)\n- ✨ [Enhancement Guide](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002Fdocs\u002FENHANCEMENT.md)\n- 🔧 [MCP Setup](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002Fdocs\u002FMCP_SETUP.md)\n\n## 📈 Stats\n\n- **16 commits** since v2.4.0\n- **700 tests** passing (up from 427, +273 new tests)\n- **4 platforms** supported (was 1)\n- **18 MCP tools** (up from 17)\n- **5 documentation guides** updated\u002Fcreated\n- **29 files changed**, 6,349 insertions(+), 253 deletions(-)\n\n## 🎉 What's New\n\nSee [CHANGELOG.md](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fblob\u002Fmain\u002FCHANGELOG.md) for complete details.\n\n## 🙏 Contributors\n\n- @yusufkaraaslan - Multi-platform architecture, all platform adaptors, comprehensive testing\n\n---\n\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fcompare\u002Fv2.4.0...v2.5.0","2025-12-28T19:44:16",{"id":254,"version":255,"summary_zh":256,"released_at":257},62364,"v2.4.0","## [2.4.0] - 2025-12-25\n\n### 🚀 MCP 2025 Upgrade - Multi-Agent Support & HTTP Transport\n\nThis **major release** upgrades the MCP infrastructure to the 2025 specification with support for 5 AI coding agents, dual transport modes (stdio + HTTP), and a complete FastMCP refactor.\n\n### 🎯 Major Features\n\n#### MCP SDK v1.25.0 Upgrade\n- **Upgraded from v1.18.0 to v1.25.0** - Latest MCP protocol specification (November 2025)\n- **FastMCP framework** - Decorator-based tool registration, 68% code reduction (2200 → 708 lines)\n- **Enhanced reliability** - Better error handling, automatic schema generation from type hints\n- **Backward compatible** - Existing v2.3.0 configurations continue to work\n\n#### Dual Transport Support\n- **stdio transport** (default) - Standard input\u002Foutput for Claude Code, VS Code + Cline\n- **HTTP transport** (new) - Server-Sent Events for Cursor, Windsurf, IntelliJ IDEA\n- **Health check endpoint** - `GET \u002Fhealth` for monitoring\n- **SSE endpoint** - `GET \u002Fsse` for real-time communication\n- **Configurable server** - `--http`, `--port`, `--host`, `--log-level` flags\n- **uvicorn-powered** - Production-ready ASGI server\n\n#### Multi-Agent Auto-Configuration\n- **5 AI agents supported**:\n  - Claude Code (stdio)\n  - Cursor (HTTP)\n  - Windsurf (HTTP)\n  - VS Code + Cline (stdio)\n  - IntelliJ IDEA (HTTP)\n- **Automatic detection** - `agent_detector.py` scans for installed agents\n- **One-command setup** - `.\u002Fsetup_mcp.sh` configures all detected agents\n- **Smart config merging** - Preserves existing MCP servers, only adds skill-seeker\n- **Automatic backups** - Timestamped backups before modifications\n- **HTTP server management** - Auto-starts HTTP server for HTTP-based agents\n\n#### Expanded Tool Suite (17 Tools)\n- **Config Tools (3)**: generate_config, list_configs, validate_config\n- **Scraping Tools (4)**: estimate_pages, scrape_docs, scrape_github, scrape_pdf\n- **Packaging Tools (3)**: package_skill, upload_skill, install_skill\n- **Splitting Tools (2)**: split_config, generate_router\n- **Source Tools (5)**: fetch_config, submit_config, add_config_source, list_config_sources, remove_config_source\n\n### Added\n\n#### Core Infrastructure\n- **`server_fastmcp.py`** (708 lines) - New FastMCP-based MCP server\n  - Decorator-based tool registration (`@safe_tool_decorator`)\n  - Modular tool architecture (5 tool modules)\n  - HTTP transport with uvicorn\n  - stdio transport (default)\n  - Comprehensive error handling\n\n- **`agent_detector.py`** (333 lines) - Multi-agent detection and configuration\n  - Detects 5 AI coding agents across platforms (Linux, macOS, Windows)\n  - Generates agent-specific config formats (JSON, XML)\n  - Auto-selects transport type (stdio vs HTTP)\n  - Cross-platform path resolution\n\n- **Tool modules** (5 modules, 1,676 total lines):\n  - `tools\u002Fconfig_tools.py` (249 lines) - Configuration management\n  - `tools\u002Fscraping_tools.py` (423 lines) - Documentation scraping\n  - `tools\u002Fpackaging_tools.py` (514 lines) - Skill packaging and upload\n  - `tools\u002Fsplitting_tools.py` (195 lines) - Config splitting and routing\n  - `tools\u002Fsource_tools.py` (295 lines) - Config source management\n\n#### Setup & Configuration\n- **`setup_mcp.sh`** (rewritten, 661 lines) - Multi-agent auto-configuration\n  - Detects installed agents automatically\n  - Offers configure all or select individual agents\n  - Manages HTTP server startup\n  - Smart config merging with existing configurations\n  - Comprehensive validation and testing\n\n- **HTTP server** - Production-ready HTTP transport\n  - Health endpoint: `\u002Fhealth`\n  - SSE endpoint: `\u002Fsse`\n  - Messages endpoint: `\u002Fmessages\u002F`\n  - CORS middleware for cross-origin requests\n  - Configurable host and port\n  - Debug logging support\n\n#### Documentation\n- **`docs\u002FMCP_SETUP.md`** (completely rewritten) - Comprehensive MCP 2025 guide\n  - Migration guide from v2.3.0\n  - Transport modes explained (stdio vs HTTP)\n  - Agent-specific configuration for all 5 agents\n  - Troubleshooting for both transports\n  - Advanced configuration (systemd, launchd services)\n\n- **`docs\u002FHTTP_TRANSPORT.md`** (434 lines, new) - HTTP transport guide\n- **`docs\u002FMULTI_AGENT_SETUP.md`** (643 lines, new) - Multi-agent setup guide\n- **`docs\u002FSETUP_QUICK_REFERENCE.md`** (387 lines, new) - Quick reference card\n- **`SUMMARY_HTTP_TRANSPORT.md`** (360 lines, new) - Technical implementation details\n- **`SUMMARY_MULTI_AGENT_SETUP.md`** (556 lines, new) - Multi-agent technical summary\n\n#### Testing\n- **`test_mcp_fastmcp.py`** (960 lines, 63 tests) - Comprehensive FastMCP server tests\n  - All 17 tools tested\n  - Error handling validation\n  - Type validation\n  - Integration workflows\n\n- **`test_server_fastmcp_http.py`** (165 lines, 6 tests) - HTTP transport tests\n  - Health check endpoint\n  - SSE endpoint\n  - CORS middleware\n  - Argument parsing\n\n- **All tests passing**: 602\u002F609 tests (99.1% pass rate)\n\n### Changed\n\n#### MCP Server Architecture\n- **Refactored to FastMCP** - Decorator-based, modular, maintainable\n- **Code reduction** - 68% smaller","2025-12-25T21:49:54",{"id":259,"version":260,"summary_zh":261,"released_at":262},62365,"v2.2.0","## [2.2.0] - 2025-12-21\n\n### 🚀 Private Config Repositories - Team Collaboration Unlocked\n\nThis major release adds **git-based config sources**, enabling teams to fetch configs from private\u002Fteam repositories in addition to the public API. This unlocks team collaboration, enterprise deployment, and custom config collections.\n\n### 🎯 Major Features\n\n#### Git-Based Config Sources (Issue [#211](https:\u002F\u002Fgithub.com\u002Fyusufkaraaslan\u002FSkill_Seekers\u002Fissues\u002F211))\n- **Multi-source config management** - Fetch from API, git URL, or named sources\n- **Private repository support** - GitHub, GitLab, Bitbucket, Gitea, and custom git servers\n- **Team collaboration** - Share configs across 3-5 person teams with version control\n- **Enterprise scale** - Support 500+ developers with priority-based resolution\n- **Secure authentication** - Environment variable tokens only (GITHUB_TOKEN, GITLAB_TOKEN, etc.)\n- **Intelligent caching** - Shallow clone (10-50x faster), auto-pull updates\n- **Offline mode** - Works with cached repos when offline\n- **Backward compatible** - Existing API-based configs work unchanged\n\n#### New MCP Tools\n- **`add_config_source`** - Register git repositories as config sources\n  - Auto-detects source type (GitHub, GitLab, etc.)\n  - Auto-selects token environment variable\n  - Priority-based resolution for multiple sources\n  - SSH URL support (auto-converts to HTTPS + token)\n\n- **`list_config_sources`** - View all registered sources\n  - Shows git URL, branch, priority, token env\n  - Filter by enabled\u002Fdisabled status\n  - Sorted by priority (lower = higher priority)\n\n- **`remove_config_source`** - Unregister sources\n  - Removes from registry (cache preserved for offline use)\n  - Helpful error messages with available sources\n\n- **Enhanced `fetch_config`** - Three modes\n  1. **Named source mode** - `fetch_config(source=\"team\", config_name=\"react-custom\")`\n  2. **Git URL mode** - `fetch_config(git_url=\"https:\u002F\u002F...\", config_name=\"react-custom\")`\n  3. **API mode** - `fetch_config(config_name=\"react\")` (unchanged)\n\n### Added\n\n#### Core Infrastructure\n- **GitConfigRepo class** (`src\u002Fskill_seekers\u002Fmcp\u002Fgit_repo.py`, 283 lines)\n  - `clone_or_pull()` - Shallow clone with auto-pull and force refresh\n  - `find_configs()` - Recursive *.json discovery (excludes .git)\n  - `get_config()` - Load config with case-insensitive matching\n  - `inject_token()` - Convert SSH to HTTPS with token authentication\n  - `validate_git_url()` - Support HTTPS, SSH, and file:\u002F\u002F URLs\n  - Comprehensive error handling (auth failures, missing repos, corrupted caches)\n\n- **SourceManager class** (`src\u002Fskill_seekers\u002Fmcp\u002Fsource_manager.py`, 260 lines)\n  - `add_source()` - Register\u002Fupdate sources with validation\n  - `get_source()` - Retrieve by name with helpful errors\n  - `list_sources()` - List all\u002Fenabled sources sorted by priority\n  - `remove_source()` - Unregister sources\n  - `update_source()` - Modify specific fields\n  - Atomic file I\u002FO (write to temp, then rename)\n  - Auto-detect token env vars from source type\n\n#### Storage & Caching\n- **Registry file**: `~\u002F.skill-seekers\u002Fsources.json`\n  - Stores source metadata (URL, branch, priority, timestamps)\n  - Version-controlled schema (v1.0)\n  - Atomic writes prevent corruption\n\n- **Cache directory**: `$SKILL_SEEKERS_CACHE_DIR` (default: `~\u002F.skill-seekers\u002Fcache\u002F`)\n  - One subdirectory per source\n  - Shallow git clones (depth=1, single-branch)\n  - Configurable via environment variable\n\n#### Documentation\n- **docs\u002FGIT_CONFIG_SOURCES.md** (800+ lines) - Comprehensive guide\n  - Quick start, architecture, authentication\n  - MCP tools reference with examples\n  - Use cases (small teams, enterprise, open source)\n  - Best practices, troubleshooting, advanced topics\n  - Complete API reference\n\n- **configs\u002Fexample-team\u002F** - Example repository for testing\n  - `react-custom.json` - Custom React config with metadata\n  - `vue-internal.json` - Internal Vue config\n  - `company-api.json` - Company API config example\n  - `README.md` - Usage guide and best practices\n  - `test_e2e.py` - End-to-end test script (7 steps, 100% passing)\n\n- **README.md** - Updated with git source examples\n  - New \"Private Config Repositories\" section in Key Features\n  - Comprehensive usage examples (quick start, team collaboration, enterprise)\n  - Supported platforms and authentication\n  - Example workflows for different team sizes\n\n### Dependencies\n- **GitPython>=3.1.40** - Git operations (clone, pull, branch switching)\n  - Replaces subprocess calls with high-level API\n  - Better error handling and cross-platform support\n\n### Testing\n- **83 new tests** (100% passing)\n  - `tests\u002Ftest_git_repo.py` (35 tests) - GitConfigRepo functionality\n    - Initialization, URL validation, token injection\n    - Clone\u002Fpull operations, config discovery, error handling\n  - `tests\u002Ftest_source_manager.py` (48 tests) - SourceManager functionality\n    - Add\u002Fget\u002Flist\u002Fremove\u002Fupdate sources\n    - Registry persistence, atomic writes, default token env\n  - `tests\u002Ftest_mcp_git_sources.","2025-12-21T20:14:01"]