Show HN: ClawVet – Offline AST scanner and rug-pull diff for AI agent skills ClawVet launched as an offline static-analysis scanner that grades AI agent skills, Cursor rules, AGENTS.md files and helper scripts for risk, using Tree-sitter taint tracking to identify which secrets, hosts and startup files the code touches. The tool runs in the browser and ships as a CLI that scans multi-file bundles, with `clawvet init` adding install-time gating and rug-pull permission diffs and a GitHub Action posting a Permissions Diff on every pull request. The release targets supply-chain risk in third-party agent skills, where a skill can change permissions after installation. A APPROVE 0/100 risk skill - Critical 0 - High 0 - Medium 0 - Low 0 0 ms taint flow proven by AST Paste a SKILL.md , Cursor rule, AGENTS.md or helper script. ClawVet runs its static rule passes plus Tree-sitter taint tracking right here in your browser and tells you which secrets, hosts and startup files the code touches. 1 Pick a sample, paste a skill or fetch a URL to see its grade, permissions and findings. The CLI scans whole multi-file bundles. clawvet init adds install-time gating with rug-pull permission diffs, and the GitHub Action posts a Permissions Diff on every PR. npm i -g clawvet && clawvet audit --summary clawvet gate --semantic