skill-creator

What This Skill Does

The skill-creator is the foundational engine for extending OpenClaw capabilities. It functions as an architectural assistant designed to help developers build, maintain, and audit modular agent skills. By providing a standardized framework for creating SKILL.md documentation and structuring bundled resources, it ensures that every new skill is consistent, efficient, and optimized for the AI's context window. This skill handles the heavy lifting of repository organization, metadata validation, and adherence to core architectural principles like maintaining low-token usage and setting appropriate degrees of freedom for agent actions.

Installation

To integrate the skill-creator into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/openclaw/skills/skill-creator

Ensure that you have appropriate write permissions for the target directory where the skill files will be generated.

Use Cases

This skill is ideal for:

• Initializing new skill repositories with valid YAML frontmatter and directory structures.
• Auditing existing skills to ensure they conform to current versioning and formatting standards.
• Refactoring messy skill directories by moving loose files into logical scripts/ or references/ subdirectories.
• Improving legacy SKILL.md files to be more concise, effectively reducing token waste and improving model performance.
• Validating that complex skills have the necessary guardrails and documentation for safe execution.

Example Prompts

1. "Create a new skill for me called 'data-scrubber' that specializes in cleaning CSV files based on specific schema requirements."
2. "Review the current SKILL.md file in my 'web-scraper' skill and improve the clarity while shortening the overall word count to save tokens."
3. "Tidy up the 'project-manager' skill directory by moving the utility bash scripts into a scripts/ folder and cleaning up stale reference files."

Tips & Limitations

• Principle of Minimization: Always prioritize the token economy. Before adding lengthy documentation, ask yourself if the AI truly needs that context for successful execution.
• Freedom Calibration: When authoring skills, choose your 'degree of freedom' carefully. High-freedom tasks rely on prompting heuristics, whereas low-freedom tasks should rely on fixed, verified scripts to prevent errors.
• Testing: Always validate your newly created skills in a sandbox environment before distribution. The skill-creator helps with structure, but the functional logic remains the author's responsibility.