audit
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill audits untrusted codebase content, which is a surface for indirect prompt injection. Mitigation is implemented via Rule 6, which directs the agent to treat repository content as data and ignore instructions like 'ignore previous instructions'. The process includes ingestion of repository files, boundary markers via instruction-override rules, capabilities limited to reading and generating reports, and sanitization guidelines for sensitive data.
- [PROMPT_INJECTION]: Documentation includes examples of prompt injection strings (e.g., 'ignore previous instructions') to guide the agent on what to ignore and report. These are identified by detectors as potential overrides but are used for defensive purposes.
- [DATA_EXFILTRATION]: Hard Rule 4 prevents the agent from copying secrets or credentials into findings or plans, requiring that only the file location and credential type be referenced.
- [COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh) to create issues, which is a safe operation involving a well-known service. The functionality is restricted by user flags and visibility checks.
Audit Metadata