skill-evals

Automate skill evaluation using PluginEval to perform static analysis, LLM-based semantic judging, and Monte Carlo simulation for Claude Code skills. It scores and benchmarks skills automatically to ensure consistent quality assurance.

Benchmark Claude Code skills by running automated evaluations that combine static analysis, LLM semantic judging, and Monte Carlo simulation. This approach generates detailed reports identifying areas for skill improvement.

Yes, you need uvx installed to run skill evaluations. The Skill depends on uvx and leverages scripts/plugin_eval to execute static analysis, LLM-based semantic judging, and Monte Carlo simulation.

Monte Carlo simulation for skill quality assurance models performance variability across multiple evaluation runs. Combined with LLM-based semantic judging and static analysis, it provides robust benchmarking and scoring for Claude Code skills.

Yes, you can generate detailed quality reports for Claude Code skills. The evaluation process outputs comprehensive documentation on skill quality, highlighting scores and identifying specific areas for improvement.