About
AI Skill Safety Scoreboard is a curated, opinionated guide to the SKILL.md files that drive AI agents. Every entry on this site runs through three stages — Test, Audit, Report — the TAR Engine pipeline behind the guide. We want a Tripadvisor for skills: independent, methodical, and unafraid to give a one-star review.
Who runs this
TAR Engine is an open-source agent audit framework. TAR stands for Test, Audit, Report — the three stages every skill goes through here. The guide you are reading is one of its showcases. We are a small team, and we audit by hand and by machine.
Submit a skill
Found a SKILL.md in the wild that deserves an audit? Open an issue on github.com/qingxuantang/tar-engine with a link and a one-line reason.
Self-host the engine
The entire audit pipeline is MIT-licensed and runs in a single container. BYOK for the semantic and adversarial passes. See the README for setup.