polish
Pass
Audited by Gen Agent Trust Hub on Jul 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to perform git operations (
git diff) and run arbitrary test suites as defined in the repository. This entails executing code from the local project environment to verify that cleanup tasks do not alter behavior.\n- [PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it ingests and analyzes untrusted data from the repository's files and git history. Maliciously crafted code or comments could attempt to deviate the agent from its intended cleanup instructions.\n - Ingestion points: Codebase content is read via
git diffin Step 1 andgrepin Step 3.\n - Boundary markers: The skill employs clear instructional boundaries using
<NON-NEGOTIABLE>tags and a segmented multi-agent architecture in Step 3 to isolate different analysis angles.\n - Capability inventory: The agent has the capability to run git commands and local test runners as part of the verification process (Steps 2, 5, and 6).\n
- Sanitization: The skill does not explicitly describe sanitization or escaping of the diff data before it is processed by the cleanup agents.
Audit Metadata