simplify
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from git diffs which could contain embedded instructions.
- Ingestion points: Phase 0 in SKILL.md reads output from git diff commands.
- Boundary markers: The skill does not define explicit delimiters or instructions to ignore commands found within the diff text.
- Capability inventory: The skill utilizes git for reading data, an Agent tool for task delegation, and file-writing capabilities to apply refactoring fixes.
- Sanitization: No evidence of sanitization or filtering of the diff content is present before processing.
- [COMMAND_EXECUTION]: The skill executes local git commands to define the scope of the code review.
- Evidence: Phase 0 in SKILL.md runs commands such as git diff @{upstream}...HEAD and git diff main...HEAD.
Audit Metadata