BENCHMARKS
Agent skill benchmark leaderboard (Claude Code)
No benchmark run has been published yet. This page will rank skills by the measured change in task pass rate once the first run is out.
What will appear here
Each benchmarked skill gets the same tasks run in Claude Code with and without the skill installed, repeated several times. The table will show pass rates for both arms, the difference with a 95% interval, how often the skill was invoked, and the median cost per run. Until then, every listed skill has a static health check.