verifying-production-changes
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest data from external sources like status endpoints, build-info routes, and logs.\n
- Ingestion points: Data is sourced from status endpoints, build-info routes, and report logs in SKILL.md.\n
- Boundary markers: None present to delimit untrusted data.\n
- Capability inventory: Shell command execution via git and network/database queries.\n
- Sanitization: No explicit sanitization of external inputs is provided.\n- [COMMAND_EXECUTION]: The skill provides a shell command snippet in SKILL.md that uses variables derived from external sources.\n
- Evidence:
git merge-base --is-ancestor \"$TARGET\" \"$DEPLOYED\".\n - Risk: Lack of input validation on variables sourced from external endpoints poses a potential command injection risk.
Audit Metadata