uw-skill-creator

You can automate Claude Code skill creation by using scriptable workflows to draft SKILL.md files, generate test prompts, and run iterative evaluation loops. This structured approach coordinates discovery, validation, and versioned improvements.

The best way to iterate on Claude Code skills is through a structured evaluation loop that captures test prompts, benchmarks results, and refines frontmatter. This workflow drives versioned improvements and optimizes trigger reliability.

Trigger optimization works by refining the skill description and frontmatter during the packaging process. This improves how accurately the skill activates when users submit relevant prompts.

Yes, pyyaml is a required dependency for running the automation scripts. It supports the parsing and drafting of frontmatter data needed to validate and scaffold SKILL.md files.

Yes, the workflow is designed to scale validation across multiple skills. It coordinates discovery, testing, and versioned improvement loops to maintain quality across a suite of Claude Code skills.

Structured interview capture is a guided framework that collects requirements and test prompts to draft initial SKILL.md files. It ensures the skill's core task and frontmatter accurately reflect the intended automation logic.