requesting-code-review
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions and the reviewer template utilize shell commands such as
git rev-parse,git log, andgit diffto identify, analyze, and stat code changes. These are standard operations for development-focused skills. - [PROMPT_INJECTION]: The
code-reviewer.mdtemplate instructs the subagent to process and analyze the output ofgit diff. This creates a surface for indirect prompt injection where malicious instructions embedded in the code being reviewed could attempt to deceive the reviewer. - Ingestion points: Code changes enter the subagent context via
{BASE_SHA}and{HEAD_SHA}parameters in thegit diffcommand incode-reviewer.md. - Boundary markers: The prompt lacks explicit delimiters or instructions to the LLM to ignore or isolate instructions found within the code diff content.
- Capability inventory: The subagent is designed to produce a technical report and assessment but does not appear to possess tools for code execution or network access.
- Sanitization: There is no evidence of sanitization or filtering of the code content before it is processed by the reviewer subagent.
Audit Metadata