ln-46-quality-gate-engineer
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes project code, manifests, and documentation, which could theoretically contain instructions intended to influence the agent's behavior.
- Ingestion points: Project source code, engineering manifests, and documentation (Checklist 1).
- Boundary markers: Instructions require the agent to distinguish effective checks from a compliant product and cite blocking instructions.
- Capability inventory: The skill can execute command-line analyzers and create small helper scripts (Checklist 2, 4).
- Sanitization: The agent is instructed to use tools for mechanical processing and keep full outputs outside of the immediate context.
- [DYNAMIC_EXECUTION]: The instructions allow for the creation and execution of small, custom scripts to implement specific engineering checks when native rules are unavailable.
- Evidence: Checklist Item 2 authorizes "small owned script only when no suitable rule exists."
- Mitigation: The skill requires isolated verification of these scripts using both positive and negative fixtures before they are considered proven (Checklist 4).
- [COMMAND_EXECUTION]: The skill utilizes the project's existing command surface to run verification tools and linters.
- Evidence: Checklist Item 4 requires running "representative allowed and forbidden fixtures through the real entrypoint."
- Constraint: The skill explicitly forbids changing delivery topology, credentials, permissions, or publishing artifacts without separate authority.
Audit Metadata