conducting-cloud-penetration-testing

SUSPICIOUS. The skill is internally coherent for authorized cloud penetration testing and uses official cloud endpoints rather than third-party exfiltration services, so it is not deceptive malware. However, it gives an AI agent explicit offensive security procedures, including credential retrieval from metadata services, privilege escalation, persistence, and disabling logging, which creates high misuse risk and exceeds what should normally be enabled for a general-purpose agent skill.