What is Skill Seekers?
Skill Seekers is the data layer for AI systems. It transforms 18 source types — documentation websites, GitHub repositories, PDFs, videos, Jupyter notebooks, Word/EPUB documents, OpenAPI specs, Confluence wikis, Notion pages, and more — into structured AI skills and RAG-ready knowledge for 12+ AI platforms.
The Problem
Building AI systems that truly understand a domain requires extensive preparation:
- AI Skill Development: 70% of time is spent preprocessing — scraping docs, analyzing code, extracting patterns
- Codebase Onboarding: Understanding a new project takes weeks of manual analysis
- AI Assistant Expertise: Generic AI responses lack deep framework and codebase knowledge
- Multi-format Needs: Different AI systems need different formats (skills, RAG, coding rules)
Result: Everyone rebuilds the same preprocessing infrastructure. Stop rebuilding. Start using.
The Solution
Skill Seekers automates AI skill creation and knowledge preprocessing:
- Extract from any of 18 source types — docs, repos, PDFs, videos, notebooks, wikis, and more
- Analyze with deep code parsing (AST analysis across 27+ languages, pattern detection, architecture mapping)
- Enhance with any AI agent — Claude, Kimi, Codex, Copilot, OpenCode, or custom agents via unified AgentClient
- Package into 12+ output platforms (AI skills, RAG pipelines, coding rules, vector DBs)
- Deploy to any AI system with one command
Result: Go from any source to production-ready AI skills in 15-45 minutes, not days.
Key Capabilities
17 Input Sources
| # | Source | CLI |
|---|---|---|
| 1 | Documentation websites | skill-seekers create <url> |
| 2 | GitHub repositories | skill-seekers create owner/repo |
| 3 | PDF documents | skill-seekers create file.pdf |
| 4 | Local codebases | skill-seekers create ./path |
| 5 | Video (YouTube/local) | skill-seekers video --url <url> |
| 6 | Word (.docx) | skill-seekers create file.docx |
| 7 | EPUB e-books | skill-seekers create book.epub |
| 8 | Jupyter Notebooks | skill-seekers create file.ipynb |
| 9 | OpenAPI/Swagger | skill-seekers create spec.yaml |
| 10 | AsciiDoc | skill-seekers create file.adoc |
| 11 | PowerPoint (.pptx) | skill-seekers create file.pptx |
| 12 | Local HTML | skill-seekers create file.html |
| 13 | RSS/Atom feeds | skill-seekers create feed.rss |
| 14 | Man pages | skill-seekers create curl.1 |
| 15 | Confluence wiki | skill-seekers confluence --space KEY |
| 16 | Notion pages | skill-seekers notion --database-id ID |
| 17 | Slack/Discord chat | skill-seekers chat --export-path dir/ |
12+ Output Platforms
| Category | Platforms |
|---|---|
| AI Skills | Claude, Gemini, OpenAI, Kimi, DeepSeek, Qwen, OpenRouter, Together, Fireworks, MiniMax, OpenCode |
| RAG/Vectors | LangChain, LlamaIndex, Pinecone, Chroma, FAISS, Haystack, Qdrant, Weaviate |
| AI Coding | Cursor, Windsurf, Cline, Continue.dev, Roo, Aider, Bolt, Kilo |
| Generic | Markdown, JSON, YAML |
Agent-Agnostic Enhancement (v3.5.0)
All enhancers now support multiple AI agents via the unified AgentClient:
# Use different AI agents for enhancement
skill-seekers create https://react.dev --agent kimi
skill-seekers create https://react.dev --agent codex
skill-seekers create https://react.dev --agent-cmd "my-custom-agent run"
# Enhancement levels control depth
skill-seekers create https://react.dev --enhance-level 2 # architecture + patternsSupported agents: Claude Code, Kimi, Codex, Copilot, OpenCode, and any custom agent.
Intelligent Processing
- Smart SPA discovery — Three-layer engine: sitemap.xml, llms.txt, and browser rendering for JavaScript sites
- Pattern recognition — Detects 10 GoF design patterns in codebases (C3.x analysis, 27+ languages + Kotlin)
- Test extraction — Extracts real usage examples from test files
- How-to generation — Creates step-by-step tutorials from workflow examples
- Prompt injection detection — Security workflow scans for injection patterns
MCP & Marketplace
- 40 MCP tools across 10 categories for Claude Code Desktop, Cursor, and other agents
- Marketplace Publisher — Publish skills to Claude Code plugin marketplace repos
- 18 agent install paths — Install skills to any supported AI coding assistant
Version
Current version: v3.5.0 (April 2026)
Latest additions:
- Grand Unification — one command, direct converters (v3.5.0)
- Agent-agnostic architecture (v3.5.0)
- Marketplace & config publishing (v3.5.0)
- Smart SPA discovery engine (v3.5.0)
- Prompt injection detection (v3.5.0)
- All content extraction features enabled by default (v3.5.0)
- 12 LLM platform targets (v3.4.0)
- 21 UML architecture diagrams (v3.4.0)
- 17 source types (v3.3.0) → 18 (v3.5.0)
- Video scraping pipeline (v3.2.0)
- 3194+ tests passing
Quick Example
# Install
pip install skill-seekers
# Create skill from any source
skill-seekers create https://react.dev --target claude
skill-seekers create facebook/react --target langchain
skill-seekers create manual.pdf --target openai
skill-seekers create notebook.ipynb --target gemini
# Use a different AI agent
skill-seekers create https://react.dev --agent kimi
# Run diagnostics
skill-seekers doctorNext Steps
- Installation Guide — Install Skill Seekers
- Your First Skill — Create your first AI skill in 3 steps
- Features Overview — Explore all capabilities
- Architecture — View 21 UML architecture diagrams
Open Source — MIT License | Community-Driven — Contributions welcome!