This is the security scanner you run before installing any skill from ClawHub, GitHub, or wherever. It chains together four different scanners (aguara for prompt injection, skill-analyzer for known malicious patterns, secrets-scan for hardcoded credentials, and structure-check for basic validation) and gives you a single verdict: BLOCKED, REVIEW, or SAFE. The output is clear about what failed and why, which matters because it never auto-installs anything. You still make the final call. If you're building a marketplace or letting an AI agent install third-party code, this is the kind of gate you want in front of that process. Dependencies are standard dev tools plus a couple specialized scanners.
npx -y skills add app-incubator-xyz/skill-vetter --skill skill-vetter --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills