一个从 GitHub 仓库生成 Skill 的 skill：搜索、分析、生成模板

GitHub to Skill Creator

Automatically discover GitHub repositories and convert them into Claude Skills.

Overview

Transforms any GitHub repository into a reusable Claude Skill by:

1. Searching GitHub for high-quality, active repositories
2. Analyzing repository structure and documentation
3. Generating skill templates with proper structure
4. Packaging ready-to-use skills

User Request
    ↓
Does user have a specific repo in mind?
    ├─ YES → Generate skill template directly
    │         (use generate_skill_template.py)
    │
    └─ NO → Search GitHub first
             1. Ask clarifying questions
             2. Run repository search
             3. Present TOP 5-10 options
             4. User selects repo
             5. Generate skill template

Phase 1: Repository Search & Recommendation

Step 1: Understand Requirements

Ask the user:

• What functionality do you need?
• Any preferred programming language?
• Specific requirements (license, activity level)?

Step 2: Search GitHub

Use the search script with appropriate filters:

# Basic search
python3 scripts/search_github_repos.py "PDF processing"

# With language filter
python3 scripts/search_github_repos.py "image manipulation" --language python

# High quality, recently updated
python3 scripts/search_github_repos.py "REST API" --min-stars 500 --sort updated

# Save results for analysis
python3 scripts/search_github_repos.py "data processing" --output results.json

Step 3: Present Results

Display results in a clear table format:

Quality Criteria

When recommending repositories, ensure:

• ✅ Active maintenance: Updated within 6 months
• ✅ Quality indicators: 100+ stars, good documentation
• ✅ Clear API: Well-defined interfaces and examples
• ✅ License: Permissive license (MIT, Apache, BSD)

Phase 2: Skill Template Generation

Step 1: Analyze Repository

After user selects a repository:

# Generate skill template
python3 scripts/generate_skill_template.py owner/repository

# Custom output directory
python3 scripts/generate_skill_template.py owner/repository --output ./skills

Step 2: Review Generated Structure

The generator creates:

skill-name/
├── SKILL.md              # Ready to customize
├── scripts/
│   └── example.py        # Starter code
└── references/           # For additional docs

Step 3: Customize the Skill

Update SKILL.md:

• Edit frontmatter description (be specific about when to use)
• Replace [TODO] items with actual content
• Add clear usage examples
• Document configuration options

Implement Core Logic:

• Write wrapper scripts in scripts/
• Handle errors gracefully
• Add logging and user feedback

Add References:

• API documentation
• Configuration examples
• Troubleshooting guides

# Test the skill
# [Manual testing steps]

# Package the skill
cd ../skill-creator
python3 scripts/package_skill.py ../skill-name

Usage Examples

Example 1: User Knows What They Want

User: "Create a skill from pdfplumber repository"

Action:

python3 scripts/generate_skill_template.py jsvine/pdfplumber

Example 2: User Needs Guidance

User: "I want to add PDF processing capabilities"

Action:

1. Ask: "What type of PDF processing? (extraction, manipulation, form filling)"
2. Search: python3 scripts/search_github_repos.py "PDF extraction" --language python
3. Present TOP 5 options with pros/cons
4. Guide selection and generate template

Example 3: No Suitable Repository Found

User: "I need quantum computing simulation"

Action:

1. Search reveals limited/old repositories
2. Suggest alternatives:

• Combine multiple smaller libraries
• Create from scratch with skill-creator
• Use web APIs instead

Best Practices

Search Strategy

• Start broad: Use general terms ("PDF processing")
• Narrow down: Add filters (--language, --min-stars)
• Try variations: "PDF extract", "PDF parse", "document processing"

Repository Evaluation

Check these before recommending:

1. Last commit: Within 6 months
2. Issues ratio: Low open/closed ratio indicates good maintenance
3. Documentation: README, examples, API docs
4. License: Prefer MIT/Apache for permissive use

Skill Design Principles

• Keep it focused: One primary purpose per skill
• Minimal dependencies: Prefer widely-used libraries
• Clear error messages: Help users troubleshoot
• Test thoroughly: Verify before packaging

Troubleshooting

No repositories found

• Broaden search terms
• Reduce minimum stars threshold
• Try different programming languages

GitHub API rate limit

• Create GitHub token: https://github.com/settings/tokens
• Set GITHUB_TOKEN environment variable
• Or use --token parameter

Generated skill doesn't work

• Verify repository has clear API/usage patterns
• Check dependencies are installable
• Review error messages in scripts

pip install requests

Optional (for GitHub API rate limits):

• GitHub personal access token

Related Skills

• skill-creator: Create skills from scratch
• pdf-to-ppt: Example of a generated skill

Resources

• GitHub Search: https://github.com/search
• GitHub API Docs: https://docs.github.com/en/rest
• Skill Creation Guide: Available in skill-creator references/