redteam
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [PROMPT_INJECTION] (SAFE): While the skill uses adversarial framing ("Think like an attacker", "break this"), these instructions are scoped to the legitimate task of analyzing a user-provided plan for weaknesses. It does not contain instructions to bypass AI safety filters or disregard system prompts.
- [NO_CODE] (SAFE): The skill does not include any Python scripts, Node.js code, or shell commands. It is purely a structured text template.
- [DATA_EXFILTRATION] (SAFE): There are no mechanisms for file access or network communication, preventing any risk of data being sent to external servers.
- [REMOTE_CODE_EXECUTION] (SAFE): No external dependencies, package managers, or remote script download patterns were identified.
Audit Metadata