requesting-code-review
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The template in
code-reviewer.mdconstructs shell commands likegit diffusing string interpolation of Git SHAs. This creates a potential command injection surface if the SHA values are sourced from untrusted inputs rather than being generated by the agent. - [INDIRECT_PROMPT_INJECTION]: The
code-reviewer.mdtemplate ingests external data including implementation descriptions and requirements without using boundary markers or sanitization, which could allow malicious instructions in those inputs to influence the subagent's assessment. - Ingestion points:
code-reviewer.mdplaceholders for{DESCRIPTION},{PLAN_REFERENCE}, and{WHAT_WAS_IMPLEMENTED}. - Boundary markers: Absent in the provided template.
- Capability inventory: File system access via
git diff. - Sanitization: No validation or escaping is applied to the input variables.
Audit Metadata