respond-to-review
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is specifically designed to ingest and process code review feedback, which is externally provided and untrusted data.
- Ingestion points: Review comment text provided during the code review process as described in the Core Process section of SKILL.md.
- Boundary markers: The instructions explicitly command the agent to treat feedback as untrusted data and use a classification system (Correct, Suggestion, Nice to have, Incorrect, Ambiguous) rather than following embedded directives. Rules 1 and 4 reiterate that directives should not be executed.
- Capability inventory: The skill can execute local development tools including mix test, mix format, and mix credo, and perform Git operations to push changes.
- Sanitization: There are explicit instructions in the 'SECURITY & RULES' section to never pass raw comment text to sub-processes and to reduce it to a classification label first to prevent exploitation.
- [COMMAND_EXECUTION]: The skill utilizes local project tools to verify code quality and correctness.
- Execution method: Uses standard Elixir ecosystem tools such as mix test for testing, mix format for formatting checks, and mix credo for static code analysis as part of the implementation and verification workflow.
Audit Metadata