Skill Creator 🛠️

Bundle plugin: Guide for creating or updating reusable OpenClaw skills.

What it does

Provides comprehensive guidance and helper scripts for creating effective OpenClaw skills:

• Best practices - Learn core principles for skill design
• Progressive disclosure - Keep skills concise and efficient
• Initialization script - Scaffold new skills from template
• Validation script - Verify skill structure and metadata
• Workflow patterns - Reference examples for common skill types

Installation

In OpenClaw

openclaw bundles install clawhub:@aisa/skill-creator

Usage

The skill loads automatically when you ask about creating or updating skills:

• "Help me create a new OpenClaw skill for [domain]"
• "Validate my skill package"
• "Improve this skill's design"

Helper Scripts

Note: Use python3 or python depending on your system.

Initialize a New Skill

python3 skills/skill-creator/scripts/init_skill.py my-new-skill

Creates a template skill at /home/ubuntu/skills/my-new-skill/

Validate an Existing Skill

python3 skills/skill-creator/scripts/quick_validate.py my-skill
python3 skills/skill-creator/scripts/quick_validate.py /home/ubuntu/skills/my-skill

(Replace python3 with python if the command is not available on your system.)

Checks:

• SKILL.md exists
• Valid YAML frontmatter
• Required fields (name, description)

Components

skill-creator/
├── .codex-plugin/
│   └── plugin.json
├── skills/
│   └── skill-creator/
│       ├── SKILL.md                 # Main guidance document
│       ├── scripts/
│       │   ├── init_skill.py        # Skill initializer
│       │   └── quick_validate.py   # Validation script
│       └── references/
│           ├── workflows.md         # Workflow pattern examples
│           ├── progressive-disclosure-patterns.md
│           └── output-patterns.md
├── openclaw.plugin.json
├── package.json
└── README.md

Skill Design Principles

1. Concise is Key

Context window is a shared resource. Only add what OpenClaw doesn't already know.

2. Appropriate Freedom Levels

• High freedom - Text instructions for flexible tasks
• Medium freedom - Pseudocode for pattern-based tasks
• Low freedom - Specific scripts for fragile operations

3. Progressive Disclosure

• Metadata - Always loaded (~100 words)
• SKILL.md body - When triggered (<500 lines)
• References - As needed

4. Skill Anatomy

skill-name/
├── SKILL.md (required)        # Metadata + instructions
└── Bundled Resources (optional)
    ├── scripts/               # Executable code
    ├── references/            # Documentation
    └── templates/             # Output assets

Examples

See SKILL.md for complete workflow patterns and examples.

License

MIT