Frequently Asked Questions

General Questions

What is Skill Seekers?

Skill Seekers is the AI Skill & RAG Toolkit. It transforms documentation websites, GitHub repositories, PDF files, and local codebases into structured AI skills and RAG-ready knowledge for:

AI Platform Skills - Claude, Gemini, OpenAI with deep expertise
AI Coding Assistants - Cursor, Windsurf, Cline, Continue.dev
RAG Pipelines - LangChain, LlamaIndex, Chroma, FAISS, Haystack, Qdrant, Weaviate
Any LLM Platform - Generic Markdown, JSON, YAML

Is it free?

Yes! Skill Seekers is 100% free and open-source (MIT License). You only pay for:

Claude API (if using API enhancement mode) - ~$0.15-$0.30 per skill
Your Claude Max subscription (if using local enhancement - recommended!)
Cloud storage (if uploading to S3/GCS/Azure)

Most features are completely free, including local AI enhancement using Claude Code.

Which platforms are supported?

Input Sources (4):

Documentation websites (any HTML docs)
GitHub repositories (public & private)
PDF files (with OCR for scanned docs)
Local codebases (27+ languages)

Output Formats (16):

AI Skills: Claude, Gemini, OpenAI
RAG/Vectors: LangChain, LlamaIndex, Chroma, FAISS, Haystack, Qdrant, Weaviate
AI Coding: Cursor, Windsurf, Cline, Continue.dev
Generic: Markdown, JSON, YAML

All features work across all platforms with complete feature parity.

Installation & Setup

What are the requirements?

Python 3.10+ (required)
pip or uv package manager
Git (for GitHub scraping)
Tesseract OCR (optional, for scanned PDFs)

How do I install?

pip install skill-seekers

That’s it! See Installation Guide for detailed instructions.

Do I need to install Claude Code?

No, but it’s highly recommended! With Claude Code, you can use FREE local AI enhancement (uses your Claude Max subscription, no API costs).

Without Claude Code, you can still:

Scrape documentation (works perfectly)
Package skills (works perfectly)
Upload manually (works perfectly)
Use API enhancement (costs ~$0.30/skill)

Usage Questions

How long does it take to create a skill?

Typical timeline:

Small docs (50 pages): 2-3 minutes
Medium docs (200 pages): 5-10 minutes
Large docs (1000+ pages): 15-30 minutes
GitHub repo analysis: 3-10 minutes
PDF extraction: 1-5 minutes
Local codebase: 5-15 minutes
AI enhancement: 30-60 seconds (local) or 10-30 seconds (API)

Total for React docs: ~12 minutes start to finish!

Can I use it without internet?

Partially:

✅ Can work offline: PDF extraction, local file analysis, local codebases
❌ Needs internet: Documentation scraping, GitHub API calls, API enhancement, upload

How accurate is the scraping?

Very accurate with proper configuration:

Smart selectors - Automatically detect content vs navigation
Category detection - 95%+ accuracy for well-structured docs
Code preservation - 100% accuracy for code blocks
Link resolution - Handles internal references correctly

Tips for best results:

Use interactive mode to test selectors
Check existing configs first (24 presets available)
Validate with skill-seekers estimate before full scrape

Feature Questions

What’s the difference between local and API enhancement?

Feature	Local (FREE)	API
Cost	FREE (uses Claude Max)	~$0.15-$0.30/skill
Speed	30-60 seconds	10-30 seconds
Quality	Same (Claude Sonnet)	Same
Requirements	Claude Code installed	ANTHROPIC_API_KEY
Use Case	Single skills, development	Batch processing, CI/CD

Recommendation: Use local mode for development, API for automation.

What are enhancement workflows (v3.1.0)?

Workflows are reusable enhancement strategies that define how AI transforms your content:

# Use a preset workflow
skill-seekers create https://react.dev --enhance-workflow security-focus

# Chain multiple workflows
skill-seekers create https://github.com/owner/repo \
  --enhance-workflow minimal \
  --enhance-workflow api-documentation

Bundled presets:

default - Balanced enhancement
minimal - Fast, light enhancement
security-focus - Security vulnerability analysis
architecture-comprehensive - Deep architectural insights
api-documentation - API-focused documentation

See Enhancement Workflows for details.

How do I create custom workflows?

# Create workflow file
cat > my-workflow.yaml << 'EOF'
name: "custom-security"
description: "Custom security analysis"
stages:
  - name: "Vulnerability Scan"
    prompt: "Analyze the code for security vulnerabilities..."
    model: "claude-sonnet-4"
    temperature: 0.2
EOF

# Add to Skill Seekers
skill-seekers workflows add my-workflow.yaml

# Use it
skill-seekers create <source> --enhance-workflow custom-security

See Custom Workflows for complete guide.

Can I scrape private documentation?

Yes! Several options:

VPN/network access - Scrape from internal network
Local HTML - Download docs locally, then scrape
Authentication - Configure custom headers/cookies
Private GitHub - Use GITHUB_TOKEN for private repos

See GitHub Analysis Tutorial for details.

How do I handle large documentation (10K+ pages)?

Use config splitting and router generation:

# Automatically split large config
skill-seekers split --config configs/large-docs.json

# Generate router skill
skill-seekers router output/large-docs-*/

This creates focused sub-skills with intelligent routing. See Large Documentation Guide for details.

Can I combine multiple sources?

Yes! Use the unified create command (v3.0+):

# Create with multiple sources
skill-seekers create https://docs.example.com \
  --github https://github.com/example/repo \
  --pdf ./manual.pdf \
  --target langchain

Or use a config file:

{
  "name": "unified-skill",
  "sources": [
    {"type": "documentation", "base_url": "https://docs.example.com"},
    {"type": "github", "repository": "example/repo"},
    {"type": "pdf", "directory": "./pdfs"}
  ]
}

See Multi-Source Tutorial for details.

Can I process local codebases?

Yes! Use the create command with a local path:

# Analyze local project
skill-seekers create ./my-project --target langchain

# Analyze Godot game with C3.x analysis
skill-seekers create ./my-game --comprehensive --target claude

Supports 27+ programming languages including Python, JavaScript, Go, Rust, C++, C#, GDScript, and more.

Platform-Specific Questions

Do I need separate skills for different AI systems?

No! Create once, package for any platform:

# Create skill (works for all sources)
skill-seekers create https://react.dev

# Package for different platforms
skill-seekers create https://react.dev --target langchain
skill-seekers create https://react.dev --target llamaindex
skill-seekers create https://react.dev --target claude

How does RAG integration work?

LangChain: Native Document objects with metadata
LlamaIndex: TextNode objects with embeddings
Vector DBs: Direct export to Chroma, Weaviate, FAISS, Qdrant
Chunking: Smart semantic chunking (512 tokens)

See RAG & Vector Databases for details.

How does AI Coding integration work?

Cursor: .cursorrules file with framework knowledge
Windsurf: .windsurfrules file
Cline: .clinerules + MCP tools
Continue.dev: HTTP context provider

All work with docs, repos, PDFs, and codebases.

Troubleshooting

Why is scraping slow?

Common causes:

Rate limiting - Increase rate_limit in config (try 1.0 or 2.0)
Sync mode - Use --async flag for 2-3x speedup
Large pages - Normal for 1000+ page sites
Network issues - Check internet connection

Why are some pages missing?

Check:

max_pages limit - Remove or increase limit
URL patterns - Add url_patterns to config
Selectors - Test with --interactive mode
JavaScript rendering - Some sites need browser automation

How do I debug selector issues?

# Interactive mode shows what gets extracted
skill-seekers create https://example.com --interactive

# Test on single page
skill-seekers estimate --config configs/test.json

See Troubleshooting Guide for more help.

Why does enhancement fail?

Common issues:

No ANTHROPIC_API_KEY - Set environment variable for API mode
Claude Code not installed - Install for local mode
Timeout - Increase --timeout flag (default: 600 seconds)
File permissions - Check write permissions on output directory

MCP Questions

What is MCP?

MCP (Model Context Protocol) is a standard for connecting AI tools. Skill Seekers provides 26 MCP tools for Claude Code Desktop, allowing natural language commands like “create a React skill”.

How do I set up MCP?

# Automatic setup (recommended)
cd /path/to/Skill_Seekers
./setup_mcp.sh

# Manual setup
# Edit ~/.claude/mcp_settings.json

See MCP Setup Guide for details.

Which AI agents support MCP?

Claude Code - stdio transport (native)
GitHub Copilot CLI - Enterprise integration
OpenAI Codex CLI - OpenAI integration
OpenCode CLI - Open source alternative
Cursor - HTTP transport
Windsurf - HTTP transport

Setup script auto-detects and configures all installed agents.

Advanced Questions

Can I customize the AI enhancement?

Yes! Enhancement uses configurable workflows. You can:

Use bundled presets (default, minimal, security-focus, etc.)
Create custom YAML workflows
Use different AI models (Claude, Gemini, GPT-4o)
Skip enhancement entirely (--skip-enhancement)
Chain multiple workflows for comprehensive analysis

Can I contribute configs?

Absolutely! We welcome community configs:

# Submit via MCP
submit_config(config_json="...", description="...")

# Or create GitHub issue
# https://github.com/yusufkaraaslan/Skill_Seekers/issues

See Contributing Guide for details.

Is there a roadmap?

Yes! See Roadmap for planned features and Changelog for version history.

Getting Help

Can’t find your answer?

📚 Documentation - Comprehensive guides
🐛 GitHub Issues - Report bugs
💬 Discussions - Ask questions
📧 Email - Direct support

Found a bug? Please report it with:

Steps to reproduce
Expected vs actual behavior
Error messages and stack traces
Environment details (OS, Python version, config file)