polish-diff
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands like
git diffand project-specific test suites to identify changes and verify that refactoring does not alter behavior. These are typical operations for a development-focused agent. - [PROMPT_INJECTION]: The skill processes code diffs which are external inputs. There is a potential for indirect prompt injection if the code contains instructions intended to mislead the cleanup agents. However, the skill implements strong guardrails including read-only agent contexts, a triage step for all findings, and a requirement that the original test suite must remain green.
- Ingestion points: Git diff output gathered in Step 1.
- Boundary markers: Explicit 'NON-NEGOTIABLE' rules regarding behavior preservation and read-only dispatch of sub-agents.
- Capability inventory: File system reading, file writing (via the test-first sub-skill), and shell execution for git and testing.
- Sanitization: Manual or controller-level triage in Step 4 and mandatory verification via existing tests in Step 6.
Audit Metadata