requesting-code-review
Pass
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local git commands to retrieve commit history and code differences for analysis. Evidence:
SKILL.mdusesgit rev-parseandgit log, whilecode-reviewer.mdinstructs the subagent to rungit diff. These operations are essential for the primary purpose of code review. - [PROMPT_INJECTION]: The subagent template in
code-reviewer.mdis susceptible to indirect prompt injection because it interpolates user-controlled data into instructions. - Ingestion points: Placeholders such as
{WHAT_WAS_IMPLEMENTED},{PLAN_OR_REQUIREMENTS}, and{DESCRIPTION}incode-reviewer.md. - Boundary markers: The template uses standard Markdown structure but lacks explicit delimiters or safety instructions to prevent the subagent from following commands embedded in the provided data.
- Capability inventory: The subagent has access to the local git repository and can execute commands to inspect code via
code-reviewer.md. - Sanitization: No sanitization or validation is applied to the content injected into the prompt placeholders.
Audit Metadata