claude-skills-security-guide

Skill integrity verifier. Checks that all installed Claude Skills match their known-good cryptographic hashes and alerts on any modifications, new unexpected files, or missing files. Use this skill to detect skill tampering, rug-pull attacks, or unauthorized post-install changes. Invoke with: "verify skill integrity", "check skill hashes", "run hash verification", or "detect skill tampering".

Script output sanitizer and second-order injection prevention. Validates and cleans text before it is passed back to Claude as context. Use this skill when processing output from untrusted scripts, web scraping results, file reads from unknown sources, or any external data that will be interpreted by Claude. Invoke with: "sanitize this output", "check this text for injection", "clean script output before processing", or "run output through the sanitizer".

Runtime security monitor for Claude Skills. Activates automatically when skills are loaded to scan all installed skill files for suspicious patterns including external URL references, credential access, persistence mechanisms, and hidden content. Run this skill before executing any untrusted skill to get a security report. Invoke with: "run security monitor", "scan installed skills", "check skills for threats", or "audit skill security".