review-and-simplify-changes
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data from git commits, PRs, and external sources.\n
- Ingestion points: Untrusted content enters the agent's context through
git diff,git log, and referenced task notes or issue descriptions (Workflow Step 3 in SKILL.md).\n - Boundary markers: The instructions provide guidelines on pinning scope but lack explicit instructions to disregard natural language commands or instructions embedded within the code diffs or commit messages.\n
- Capability inventory: The agent is authorized to execute shell commands (git, knip, madge) and read/write to the file system (Workflow Step 2 and 7 in SKILL.md).\n
- Sanitization: No explicit sanitization or filtering of external content is defined before the agent processes the diff information.\n- [COMMAND_EXECUTION]: The skill utilizes shell commands for repository analysis and tool invocation.\n
- Evidence: The skill explicitly instructs the agent to use
git status,git diff,git log, and static analysis tools such asknipandmadgeto identify issues (Workflow Step 2 and Eight-Topic Coverage Checklist in SKILL.md).
Audit Metadata