skill-creator

Define what the skill should do and when it should trigger, then write a SKILL.md with name and description frontmatter plus markdown instructions. Add scripts, references, or assets directories for deterministic tasks, detailed docs, or output templates.

Run each test prompt twice in parallel: once with the skill and once without as a baseline. Grade both outputs against objective assertions, then aggregate pass rates, timing, and token usage into a benchmark comparison.

Generate 20 realistic queries split between should-trigger and should-not-trigger cases, then run the optimization loop. It evaluates description variants on a train split and selects the best by held-out test score to avoid overfitting.

Include both what the skill does and specific contexts for when to use it, since all triggering information lives in the description. Slightly assertive phrasing helps because Claude tends to undertrigger skills on borderline queries.

Skills with objectively verifiable outputs like file transforms, data extraction, or code generation benefit from quantitative assertions. Skills with subjective outputs like writing style are better evaluated through qualitative human review.

Claude.ai lacks subagents, so test runs execute sequentially without baseline comparisons or quantitative benchmarking. Description optimization requires the claude CLI and is unavailable; evaluation relies on inline qualitative feedback.