requesting-code-review
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The subagent template in
code-reviewer.mdingests external data (descriptions and requirements) via placeholders without explicit boundary markers or sanitization logic. - Ingestion points: The
{DESCRIPTION}and{PLAN_REFERENCE}placeholders incode-reviewer.md. - Boundary markers: Absent. There are no delimiters or instructions provided to the subagent to treat placeholder content as untrusted data.
- Capability inventory: The agent can execute git commands and its feedback directly influences the production readiness assessment and subsequent development steps.
- Sanitization: Absent. There is no evidence of input validation or escaping for the content interpolated into the prompt.
- [COMMAND_EXECUTION]: The
code-reviewer.mdtemplate defines shell commands that incorporate variables without strict validation. - Evidence: The command
git diff {BASE_SHA}..{HEAD_SHA}is executed within the subagent context. - Risk: If the input SHAs are not strictly validated as hexadecimal strings, an attacker could inject shell metacharacters to perform unauthorized command execution.
Audit Metadata