Skill Security Auditor

**Judgement:** Strongly recommended as a *first-line*, pattern-based security check for ClawHub skills, especially in light of campaigns like ClawHavoc. It materially improves safety but must not be treated as a complete security solution. **What it does well:** - Provides a local Bash script (`analyze-skill.sh`) that never needs to execute the target skill, only read its SKILL.md and metadata. - Detects a broad range of high-signal indicators: fake prerequisite installers, `curl | bash`-style execution, hardcoded secrets, suspicious dependencies, known C2 IPs/domains (including ClawHavoc IOCs), and obfuscation patterns. - Assigns a 0–100 risk score with clear qualitative levels (SAFE/LOW/MEDIUM/HIGH/CRITICAL) and explicit recommendations (INSTALL / DO NOT INSTALL / REVIEW MANUALLY). - Supports both single-skill audits (by slug or file) and batch scanning of all installed skills, making it practical for routine hygiene. - Documentation is detailed and security-oriented, including guidance on false positives, VirusTotal cross-checking, and defense-in-depth. **Key risks / limitations:** - Detection is pattern/heuristic-based; sophisticated or novel malware that does not match known patterns can evade it. - False positives are possible, especially for advanced or unusual skills that legitimately use binaries, networking, or encoding techniques. - It is advisory only: it does not enforce blocking or auto-scan; users must remember to run it and still apply their own judgment. - Threat intel updates (IOC database) are manual (`curl` into the patterns directory); if users do not update, coverage against new campaigns will degrade. **Recommended scenarios:** - Always run this tool before installing any untrusted or newly published skill, particularly those touching wallets, credentials, or system-level actions. - Use it in moderation / security workflows to quickly triage suspicious skills flagged by users or automated systems. - Integrate it (via scripts) into pre-merge or pre-release checks for skill repositories, as a guardrail against accidental credential leaks or obviously risky patterns. Used as part of a broader workflow (VirusTotal checks, manual review, and platform trust signals), this skill can significantly reduce exposure to known-bad and low-effort malicious skills while remaining low-risk to run itself.