security-review

SUSPICIOUS but not malicious. The skill is internally coherent and uses official tooling, so supply-chain and credential-routing risk are low. However, it materially expands an AI agent into a security scanning/exploit-validation tool, can act on untrusted code and PR comments, can install trust transitively via another skill, and can autonomously write comments/branches/reports. This makes it high-risk operationally even though the stated purpose matches the capabilities.