skill-creator

Benchmark Claude Code skills by running quantitative evaluation loops that grade outputs against assertions and measure pass rate, timing, and token usage against baseline performance. The skill-creator automates this with aggregation scripts that generate HTML reports.

Skill description optimization iteratively improves SKILL.md triggering accuracy using train/test splits and automated optimization loops. You need it when skills undertrigger or produce inconsistent results during evaluation.

Create a Claude Code skill by capturing intent, writing a SKILL.md file, and organizing bundled resources like scripts and references. The skill-creator guides this workflow to ensure reliable triggering and consistent performance.

Yes, you need Python to run skill evaluation workflows, execute aggregation scripts, and generate HTML benchmarking reports. The skill-creator requires Python for running evaluation loops and quantitative performance analysis.

AI skills undertrigger or produce inconsistent results without structured evaluation and description optimization. The skill-creator addresses this through quantitative benchmarking, baseline comparisons, and iterative instruction improvements based on automated feedback.