Agent Security Audit
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists of documentation and auditing templates designed to improve agent security. It contains no executable malicious code, obfuscation, or persistence mechanisms.
- [PROMPT_INJECTION]: The skill functions by analyzing external codebases, which constitutes an indirect prompt injection surface. 1. Ingestion points: Target codebase processed during the audit workflow. 2. Boundary markers: None defined in the audit template to separate target data from audit instructions. 3. Capability inventory: The skill's checklist refers to tool execution (eval, exec) and file system access, which may be available to the agent performing the audit. 4. Sanitization: The checklist itself (Category A02) requires the auditor to verify sanitization in the target system, showing a defensive design despite the ingestion surface.
Audit Metadata