# Skill Seekers

> Open-source CLI and MCP server that converts docs, repos, PDFs, videos and other sources into AI skills and RAG-ready knowledge.

Skill Seekers is an open-source toolkit by Yusuf Karaaslan that turns documentation sites, GitHub repositories, PDFs, videos, notebooks, wikis and other sources into structured knowledge for AI systems. The project describes 18 source types and 22 export targets, spanning AI skills, RAG pipelines and AI coding assistants. It is installed from PyPI and driven by a command-line interface, with an MCP server also available.

## What It Is

Skill Seekers is a data-preparation layer for AI. You point a unified `create` command at a source such as a docs URL, a GitHub repo, a local codebase, a PDF, a Word file, EPUB, Jupyter notebook, OpenAPI spec, Confluence space, Notion database or Slack/Discord export. It then extracts and categorizes the content, optionally enhances it with an AI agent, and packages the result for a chosen target. For Claude, this yields a skill ZIP with a `SKILL.md` and reference files; for RAG it yields chunked documents with metadata.

## Workflow and Outputs

The documented pipeline is scrape, categorize, enhance, package and optionally upload. Documentation scraping checks `llms.txt` first and uses sitemap and headless browser rendering for JavaScript SPA sites. GitHub and local code analysis (C3.x) uses AST parsing, design pattern detection, test example extraction, how-to guide generation and architecture overviews. Video input supports transcripts and OCR of on-screen code. Exports include Claude, Gemini, OpenAI, Kimi, DeepSeek, Qwen and other LLM platforms, plus LangChain, LlamaIndex, Haystack, Chroma, FAISS, Weaviate, Qdrant and Pinecone formats, and Markdown/JSON/YAML.

## Agents, MCP and Installation to Tools

Enhancement runs through one agent client, in API mode or local mode with agents such as Claude Code, Kimi Code, Codex, Copilot and OpenCode, with workflow presets that can be chained or customized. The MCP server exposes 40 tools for Claude Code, Cursor, Windsurf, VS Code with Cline and IntelliJ IDEA. An `install-agent` command places generated skills into 19 coding agents' skill directories. An AI-driven `scan` command reads a project's manifests and imports and emits one config per detected framework.

## Setup Path

Install with `pip install skill-seekers` (Python 3.10+ and Git required), `uv tool install`, or from source. Optional extras add LLM platforms, MCP, video, Jupyter, PowerPoint, Confluence, Notion, RSS and chat export support. A `doctor` command runs diagnostic checks on the installation and environment.

## Features
- 18 source types: docs, GitHub repos, PDFs, videos, notebooks, wikis and more
- 22 export targets for AI skills, RAG/vector stores and coding assistants
- C3.x codebase analysis with AST parsing and design pattern detection
- SPA documentation discovery via sitemap, llms.txt and browser rendering
- AI enhancement with chainable workflow presets and agent-agnostic client
- MCP server with 40 tools
- Install skills into 19 AI coding agents
- AI-driven project scan generating per-framework configs
- Video transcript and on-screen code OCR extraction
- Prompt injection detection workflow
- Optional SQLite FTS5 search index in generated skills
- Doctor diagnostic command
- Multi-source unified scraping with conflict detection

## Integrations
Claude, Gemini, OpenAI, Kimi, DeepSeek, Qwen, OpenRouter, LangChain, LlamaIndex, Pinecone, Chroma, FAISS, Haystack, Qdrant, Weaviate, Cursor, Windsurf, Cline, Continue.dev, Aider, Confluence, Notion, Slack, Discord, GitHub Actions

## Platforms
MACOS, API, VSC_EXTENSION, JETBRAINS_PLUGIN, DEVELOPER_SDK, CLI

## Pricing
Open Source

## Version
v3.10.0

## Links
- Website: https://skillseekersweb.com/
- Documentation: https://skillseekersweb.com/docs/getting-started/overview
- Repository: https://github.com/yusufkaraaslan/Skill_Seekers
- EveryDev.ai: https://www.everydev.ai/tools/skill-seekers