attacking-entra-id

SUSPICIOUS/HIGH-RISK skill. Its capabilities are internally consistent with its stated pentesting purpose, so it is not deceptive malware, but it materially equips an AI agent to perform offensive cloud identity attacks, credential theft, token replay, and hybrid compromise. No hidden installer, covert exfiltration endpoint, or obfuscation is present, yet the operational impact is high enough to classify as a dangerous security skill.