skill-creator

Create a Claude skill by drafting a SKILL.md file with frontmatter and instructions, then run evals and benchmarks to measure trigger accuracy and iterate until performance is reliable.

A SKILL.md file uses structured frontmatter and an instructional body to define trigger conditions and execution instructions, enabling Claude to accurately identify when to apply the skill.

Run evals and benchmarks through a repeatable loop that tests trigger accuracy and reliability across iterations, collects metrics, and generates reports to guide prompt optimization.

Yes, pyyaml is a required dependency for the Skill Creator, used to parse and manage the structured YAML frontmatter in SKILL.md files during the creation and iteration workflow.

The best way to optimize skill descriptions is to iterate through a loop of running evals, reviewing benchmark reports, and refining SKILL.md content and test prompts based on collected performance metrics.

Unreliable skill triggering usually stems from poorly defined SKILL.md descriptions, which you can fix by running benchmarks, analyzing metric reports, and iterating on trigger instructions until accuracy improves.