requesting-code-review
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local
gitcommands (rev-parse,log,diff) to identify and display code changes. These are standard operations for a code review workflow and use the agent's provided tooling. - [INDIRECT_PROMPT_INJECTION]: The skill processes external code content which could contain malicious instructions.
- Ingestion points: The subagent reads project code via
git diffoutput as defined incode-reviewer.md. - Boundary markers: The instructions do not currently include explicit delimiters or warnings to the subagent to ignore instructions embedded in code comments or strings.
- Capability inventory: The reviewer subagent analyzes text and uses the
gittool to read repository data. - Sanitization: The code content is passed directly to the subagent without sanitization or filtering of potential injection patterns.
Audit Metadata