review-implementation
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code diffs and is instructed to 'run a narrow non-destructive check' and 'Resolve actual available skill paths'. If a malicious diff contains instructions disguised as code or comments, the agent might be tricked into executing unintentional commands or accessing unauthorized file paths during the review process.
- Ingestion points: The skill ingests untrusted code diffs and directly affected file paths (SKILL.md).
- Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from following instructions embedded within the diff content.
- Capability inventory: The skill allows for running checks (potentially executing scripts/tests) and resolving file paths, which may involve platform tools for file system access.
- Sanitization: No sanitization or validation of the input diff content is described.
- [COMMAND_EXECUTION]: The instructions explicitly direct the agent to 'run a narrow non-destructive check' to verify material claims. This permits the agent to invoke tools or shell commands to validate the code being reviewed, which could be exploited if the agent context is not strictly isolated.
Audit Metadata