red-team-review
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security vulnerabilities were identified in the analyzed files. The skill's primary function is to enhance the security and validity of other artifacts through adversarial scrutiny.
- [PROMPT_INJECTION]: The skill instructions and evaluation cases were analyzed for injection risks. The agent is explicitly instructed to refuse manual overrides of negative verdicts, and the evaluation suite confirms the agent should block attempts to bypass the 'Approved' requirement.
- [DATA_EXFILTRATION]: No unauthorized network operations or credential exposures were found. The skill uses a 'manifest.json' to explicitly define which files are included in review packets, reducing the risk of accidental sensitive data exposure.
- [COMMAND_EXECUTION]: While the skill permits the use of the 'Bash' tool, its usage is described for directory isolation and context management. No patterns of privilege escalation, persistence, or dangerous shell commands were detected.
Audit Metadata