reverse-engineer

SUSPICIOUS. The skill is internally consistent and does not show credential theft, hidden data flows, or suspicious installers, but it gives an AI agent high-risk offensive security/reverse-engineering capability. This makes it dangerous in scope even without signs of malware.