Security scanner for OpenClaw skills.
Scan before you trust.
Finds hidden instructions, HTML comments, unicode tricks, and identity overrides in SKILL.md
Detects base64+POST patterns, reverse shells, connections to webhook.site, RequestBin, and more
eval(), exec(), subprocess, pickle.loads, dynamic imports: everything that enables code execution
Path traversal, SSH key access, /etc/passwd reads, OpenClaw config with your secrets
Extracts all URLs, classifies APIs (OpenAI, Slack, GitHub...), finds hardcoded IPs
Scans automatically after every clawhub install. You know instantly if a skill is safe.
"Install the crypto-tools skill"
clawhub install crypto-tools
skill-audit checks all files immediately
🟢 Safe / ⛔ Dangerous — user decides
Real skills from ClawHub + a malicious test skill