Skill Scanner

Scan Clawdbot and MCP skills for malware, spyware, crypto-miners, and malicious code patterns before you install them. Security audit tool that detects data exfiltration, system modification attempts, backdoors, and obfuscation techniques.

This package appears to implement a local static scanner and a Streamlit UI that scans only the files you provide. Before installing or running it, do the following: 1) Verify the source/author — the registry metadata shows no homepage and the origin is unknown; prefer code from a trusted repo. 2) Inspect the full skill_scanner.py and streamlit_ui.py (the provided copy was truncated in places) to confirm there is no hidden behavior (network calls, code execution, auto-update). 3) Do not point the scanner at real secret stores or upload sensitive files to the web UI — it only looks for strings/patterns in files, but uploading sensitive data to a web UI increases exposure. 4) Run it in a sandbox or VM first and test on harmless sample skills to validate false-positive/negative behavior. 5) Note minor implementation issues (the UI references a format_markdown method and truncated code made it impossible to confirm all functions) — fix or review those before relying on automated CI gating. If you want higher assurance, ask the publisher for a canonical repo URL, full source, and a reproducible build or have a security-savvy reviewer audit the complete code.

Type: OpenClaw Skill Name: skill-scanner Version: 0.1.2 The OpenClaw AgentSkills skill bundle 'skill-scanner' is classified as benign. Its explicit purpose, as stated in SKILL.md and README.md, is to act as a security audit tool for other skills, detecting malicious patterns like data exfiltration, system modification, and arbitrary code execution. The core logic in `skill_scanner.py` implements this by scanning files against a comprehensive list of `THREAT_PATTERNS`. These patterns are indeed indicators of malicious behavior, but the scanner's code is designed to *detect* them, not *perform* them. There is no evidence of the skill itself engaging in any harmful activities, prompt injection against the agent (beyond its legitimate function), or obfuscation of its own intent.