cloud-penetration-testing

High-risk offensive security skill. Its actions align with its stated pentesting purpose, but that purpose includes credential harvesting, secret extraction, persistence, and privilege escalation, making it dangerous for an AI agent even without obvious deception or third-party exfiltration.

This code fragment is not benign infrastructure automation; it is an offensive cloud pentesting toolkit that includes direct credential-access (password spraying with persistence of valid credentials), token acquisition (OAuth device-code and IMDS managed-identity token retrieval), and a tenant privilege-escalation workflow (Graph role assignment). It also performs broad sensitive configuration/data harvesting into local files. Even without obfuscation, its operational capabilities present a high security risk and strong misuse potential.