Is that AI skill safe to run?
People download agent skills from all over the internet and run them with full access to their files and accounts. Check one first: paste the skill or a GitHub link and get a safety grade in about 20 seconds.
How it works
Paste or link the skill
Drop in the full SKILL.md, agent instructions, or a public GitHub repo, file, or gist link.
We scan and review it
A pattern scanner flags risky commands, then an AI reviewer reads the whole skill in context: injection tricks, data leaks, destructive actions, and over-broad permissions.
Get a grade you can act on
A safety score, an A+ to F grade, the concrete risks found, and a plain-English read on what the skill is useful for and where it does not belong.
Every free account gets 2 checks per day. The report is an automated review to help you decide, not a certification: for anything that touches customer data or money, read the skill yourself too.