Contributing to Claude AI Research Skills

Contributing to Claude AI Research Skills

Thank you for your interest in contributing! This guide will help you add new skills to the library.

---

🎯 What We're Building

**Vision**: The most comprehensive open-source library of AI research skills for Claude Code.

**Target**: 86 comprehensive skills covering the entire AI research lifecycle — from ideation to paper writing. ✅ Achieved.

**Current Progress**: 86/86 skills across 22 categories (100%)

**Philosophy**: Quality > Quantity. We deleted 9 low-quality skills to maintain high standards.

---

🤝 How to Contribute

Ways to Contribute

undefined

---

📝 Adding a New Skill

Step 1: Choose a Skill

Step 2: Fork and Clone

# Fork the repository on GitHub first
git clone https://github.com/YOUR_USERNAME/AI-research-SKILLs.git
cd claude-ai-research-skills

# Create a feature branch
git checkout -b add-vllm-skill

Step 3: Use Skill Seeker MCP

**Option A: Documentation Scraping**

# Create config file
python3 cli/doc_scraper.py --interactive
# Or copy and modify an existing config
cp configs/react.json configs/vllm.json

# Scrape and build
python3 cli/doc_scraper.py --config configs/vllm.json

**Option B: GitHub Scraping**

# Scrape from GitHub repository
export GITHUB_TOKEN=$(gh auth token)
python3 cli/github_scraper.py --repo vllm-project/vllm --name vllm --description "High-performance LLM inference with PagedAttention"

**Option C: Unified Scraping** (recommended for comprehensive skills)

# Combine documentation + GitHub + PDF
python3 cli/unified_scraper.py --config configs/vllm_unified.json

Step 4: Move to Correct Directory

# Determine the category (see directory structure below)
mv output/vllm/ 12-inference-serving/vllm/

# Move metadata
mv output/vllm_data/ .metadata/vllm_data/

Step 5: Validate Quality

**Based on [Anthropic Official Best Practices](anthropic_official_docs/best_practices.md)**

**Core Requirements** (or skill will be rejected):

undefined

**Gold Standard** (aim for this):

undefined