Agent skills, tested before you install them.
There are millions of SKILL.md files out there and no way to tell the good ones from the risky ones. We run each skill on a standard task, with and without it loaded, scan it for security red lines, and publish the score — evidence included.
Latest deep test
Browse by task
Development Workflow
Planning, code review, git and day-to-day engineering process.
Testing & QA
Test generation, browser testing, debugging and verification.
Frontend & Design
UI design, styling, layout and visual identity for the web.
Documents & Office
Word, PDF, spreadsheets, slides and document automation.
Data Analysis
Cleaning, querying, charting and explaining data.
DevOps & Infra
CI/CD, cloud, containers, monitoring and release work.
Marketing & Content
Copywriting, SEO content, social and growth workflows.
Security
Code audit, hardening, scanning and safe automation.
How scoring works
Two layers. First, six security red lines — fail one and the skill is flagged, whatever its score. Then 100 points across six dimensions, with live test results worth 30: the only way to earn them is to beat the no-skill baseline on a standard task. Download counts and GitHub stars are shown as background and never count toward the score. Read the full methodology →