Anthropic Skill Builder for OpenClaw

Create production-quality Claude/OpenClaw skills following Anthropic's official design patterns.

This skill helps you design, structure, and refactor OpenClaw skills using Progressive Disclosure principles from Anthropic's Complete Guide to Building Skills for Claude.

🎯 What This Skill Does

Anthropic Skill Builder guides you through creating well-structured skills that:

• ✅ Minimize context usage — Only load what Claude needs, when it needs it
• ✅ Follow official patterns — Proven design patterns from Anthropic
• ✅ Scale gracefully — Handle simple and complex use cases efficiently
• ✅ Are maintainable — Clear structure, easy to update

🚀 Quick Start

Installation

# For OpenClaw users
cd ~/.openclaw/skills
git clone https://github.com/llh315434220/anthropic-skill-builder.git

# Restart OpenClaw gateway
openclaw gateway restart

Usage

Once installed, the skill triggers automatically when you ask about:

• Creating new skills
• Improving existing skills
• Refactoring skill structure
• Applying progressive disclosure patterns

Example prompts:

"Help me create a skill for PDF processing"
"Refactor my database-query skill to use progressive disclosure"
"My skill is too long, how do I split it into references?"

📖 Core Concepts

Progressive Disclosure (3 Levels)

Skills should reveal information progressively, not dump everything into context at once:

1. Metadata (~100 words, always loaded)

   • Skill name + comprehensive description
   • All triggering logic goes here

2. SKILL.md (<500 lines, loaded when triggered)

   • Core workflow and navigation
   • Links to detailed references

3. Resources (loaded as needed)

   • scripts/ — Executable code
   • references/ — Detailed documentation
   • assets/ — Templates, images, etc.

The Golden Rule

The context window is a public good.

Every token counts. Ask: "Does Claude really need this?"

Skills should provide only non-obvious procedural knowledge, not general capabilities Claude already has.

🏗️ Skill Structure

my-skill/
├── SKILL.md                 # Main skill file (< 500 lines)
├── scripts/                 # Executable scripts (avoid rewriting fragile code)
│   ├── example.py
│   └── helper.sh
├── references/              # Detailed docs (loaded on demand)
│   ├── API_REFERENCE.md
│   ├── ADVANCED_PATTERNS.md
│   └── TROUBLESHOOTING.md
└── assets/                  # Templates, images, config files
    ├── template.json
    └── icon.png

📚 Design Patterns

Pattern 1: Multi-Framework Skill

For skills supporting multiple frameworks/domains (AWS/GCP/Azure):

cloud-deploy/
├── SKILL.md              # Overview + framework selection
└── references/
    ├── aws.md            # AWS-specific details
    ├── gcp.md            # GCP-specific details
    └── azure.md          # Azure-specific details

Pattern 2: Basic + Advanced Skill

Common operations in SKILL.md, complex cases in references:

pdf-processing/
├── SKILL.md              # Common operations
└── references/
    ├── FORMS.md          # Form filling
    ├── ENCRYPTION.md     # Password protection
    └── OCR.md            # Text extraction

Pattern 3: Database Query Skill

Split schemas by domain to avoid loading irrelevant tables:

database-analytics/
├── SKILL.md              # Query patterns
└── references/
    ├── finance_schema.md
    ├── sales_schema.md
    └── product_schema.md

See more patterns in references/PATTERNS.md

🛠️ Workflow

1. Understand — Gather 3-5 concrete examples
2. Plan — Identify scripts, references, assets needed
3. Structure — Create directory layout
4. Write SKILL.md — Core workflow + navigation
5. Create references — Split domain-specific details
6. Add scripts — Package fragile/repetitive code
7. Test — Validate all scripts work

📋 Best Practices

✅ Do:

• Put ALL triggering logic in description (not body)
• Keep SKILL.md under 500 lines
• Split large skills into references/
• Use imperative voice: "Read X", not "You can read X"
• Test all scripts before committing

❌ Don't:

• Create README.md, CHANGELOG.md in skill directory (meta-docs belong in repo root)
• Exceed 500 lines in SKILL.md without splitting
• Explain things Claude already knows
• Nest references deeper than one level

📖 Documentation

• SKILL.md — Core skill implementation
• PATTERNS.md — Detailed design patterns with examples
• QUICK_REFERENCE.md — Cheat sheet for common tasks
• USAGE.md — Detailed usage guide with examples

🤝 Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-pattern)
3. Commit your changes (git commit -m 'Add amazing pattern')
4. Push to the branch (git push origin feature/amazing-pattern)
5. Open a Pull Request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Anthropic for the Complete Guide to Building Skills for Claude
• OpenClaw team for the extensible skill system
• All contributors who help improve skill design patterns

🔗 Resources

• Anthropic Official Guide (PDF)
• OpenClaw Documentation
• OpenClaw GitHub
• Skill Creator Skill

📧 Support

• Issues: GitHub Issues
• Discord: OpenClaw Community
• Documentation: OpenClaw Docs

---

Made with ❤️ for the OpenClaw community