ai-code-cleanup
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates on code from feature branches, which acts as an ingestion point for untrusted data. The instructions explicitly direct the agent to remove 'defensive bloat,' such as null checks and error handling, potentially weakening security controls if necessary protections are misidentified as redundant.
- Ingestion points: File content is read based on the output of
git diffacross the branch history. - Boundary markers: There are no explicit markers or instructions to isolate the code being analyzed from the agent's instructions.
- Capability inventory: The skill possesses file-write capabilities and is instructed to execute shell commands to run tests.
- Sanitization: No sanitization or security validation of the branch code is performed before it is processed or tested.
- [COMMAND_EXECUTION]: The skill employs shell commands to identify modified files (
git diff) and to verify changes by running the project's test suite. This results in the execution of potentially untrusted code from the branch during the verification phase.
Audit Metadata