agentic-pentest
Audited by Socket on Sep 19, 2026
2 alerts found:
Securityx2SUSPICIOUS: the skill’s offensive capabilities are aligned with its stated purpose, but that purpose is to operationalize AI-driven penetration testing and exploitation. It enables autonomous command execution, offensive tool orchestration, credential-related workflows, evasion techniques, and third-party framework installs, making it a high-risk security skill even without clear evidence of hidden credential theft or outright malware.
The fragment documents and implements an offensive-security automation framework with legitimate purple-team and pentesting use cases. It is not clearly malware, but it contains high-risk execution patterns: LLM- or input-derived shell commands executed with shell=True, weak scope enforcement, potentially sensitive data sent to hosted LLMs, exposed Docker services, and mutable container images. These issues warrant security review and hardening before use, especially safe subprocess invocation, structured command allowlisting, robust target validation, secret redaction, authenticated services, and pinned image versions.