security-audit
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the analysis of untrusted source code from external repositories, which represents an indirect prompt injection surface. A malicious codebase could contain instructions disguised as comments or data intended to subvert the behavior of the auditing agents.
- Ingestion points: Agents (scouts and hunters) are directed to read target files as defined in 'references/reconnaissance.md' and 'references/hunting.md'.
- Boundary markers: The instructions provide workflow stages but do not define specialized escape markers or delimiters for ingested content.
- Capability inventory: The skill utilizes 'scripts/audit_run.py' for repository metadata collection and 'scripts/promote_evidence.py' for file system management.
- Sanitization: No explicit sanitization or filtering of the target source code is performed before it enters the agent context.
- [COMMAND_EXECUTION]: The script 'scripts/audit_run.py' executes the 'git' command using the 'subprocess.run' module to gather repository information. This is implemented securely using an argument list with no shell-level interpretation, which significantly reduces the risk of command injection.
Audit Metadata