Open-source skill benchmarking

Benchmark Any Skill. Prove It Works.

Submit your Claude Code skills, let AI generate rigorous benchmarks, and see how they perform across multiple models and agent loops. Data-driven quality for the skill ecosystem.

100+Skills Benchmarked

5AI Models

3Agent Loops

How It Works

Four steps from submission to score

Step 1

Submit GitHub URL

Point us to your SKILL.md file or repository. We parse skill metadata automatically.

Step 2

AI Generates Benchmarks

Our system creates targeted test scenarios that probe the skill's capabilities and edge cases.

Step 3

Multi-Agent Testing

Multiple AI models run your skill through structured agent loops, scoring each execution.

Step 4

Get Your Score

View detailed results, model-by-model breakdowns, and your position on the leaderboard.

Featured on the Leaderboard

Top-performing skills ranked by composite benchmark score

#	Skill	Author	Category	Models	Score	Trend
	commit-wizard	austindixson	Git	5	94.2	+2.1
	test-generator-pro	devcraft	Testing	5	91.8	+0.4
	refactor-engine	codesmith	Refactoring	4	89.5	-0.3
4	doc-autowriter	opensrc	Docs	5	87.1	+1.7
5	debug-detective	bugslayer	Debugging	3	85.6	+0.9

Ready to benchmark your skill?

Join the community of developers building and testing Claude Code skills. Submit yours today and see how it stacks up.