code-review-agentteam
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from git repositories which could contain adversarial instructions targeting the LLM.
- Ingestion points: Ingests data from
git diff,git log, and files indocs/,specs/, or.scratch/folders (SKILL.md). - Boundary markers: No explicit delimiters or warnings are used to instruct the agents to disregard instructions embedded within the diff or commit messages.
- Capability inventory: Teammate agents are granted
editandwritetools, creating a risk if they are manipulated into modifying the repository instead of just reviewing it (SKILL.md). - Sanitization: There is no logic to sanitize or escape the repository content before it is processed by the Reviewer agents.
Audit Metadata