1. Vet at install
A tiny hook sits in your agent. Every skill or MCP server install is checked before it runs; a vet-skills.lock pins the exact bytes that were vetted. Warn, block, or allow by policy.
Install the CLI →KillerSkills checks every agent skill and MCP server at install time, runs it in a honeypot, and tells you what it actually did. Free in your agent. Governed for your company.
A tiny hook sits in your agent. Every skill or MCP server install is checked before it runs; a vet-skills.lock pins the exact bytes that were vetted. Warn, block, or allow by policy.
Install the CLI →Every release runs in a sandbox with decoy credentials, a watched MEMORY.md and a recording proxy, usually within an hour of being published. The verdict is a behavior trace, not a guess, and the method is public.
Read the methodology →Every verdict is a signed attestation you can check offline against our published key. vet-skills verify does it in one command; CI can refuse anything whose bytes changed since review.
Signed verdicts →We diff every release and re-detonate. The hook compares what a skill does in your real sessions with what the sandbox saw. Anything new is a drift alert, and anyone can watch a skill for changes.
See drift alerts →Every public skill and MCP server on skills.sh, ClawHub, the Claude Code plugin marketplaces, npm, PyPI, the Docker MCP Catalog and the official MCP Registry, continuously re-scanned. Every verdict has a permanent page.
Browse all skills →See every skill and MCP server installed across the company, enforce an allow-list from the install hook, and get an audit trail your compliance team will accept.