review-change
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides instructional guidelines for an AI agent to act as a code reviewer. It does not include any scripts, shell commands, or network operations. All instructions are within the expected scope of code quality assessment and regression testing.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external untrusted data (code diffs and task briefs), creating a potential attack surface for indirect prompt injection. However, as this is the primary purpose of a code review tool and no specific exploitation capabilities are present within the skill itself, the risk is negligible.
- Ingestion points: Processes diffs, branch names, patch files, and project briefs provided in the environment or by the user.
- Boundary markers: The instructions encourage the agent to 'reread the diff from scratch' and maintain an independent perspective, though no formal delimiters are defined for the input data.
- Capability inventory: No internal capabilities for file system writes or network access are specified; the skill primarily references other instructional guidelines (/tdd, /architecture, etc.).
- Sanitization: No explicit sanitization of input text is performed.
Audit Metadata