requesting-code-review

Pass

Audited by Gen Agent Trust Hub on May 7, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes task descriptions and implementation requirements which are interpolated into a subagent's prompt without protective delimiters or sanitization. This creates a surface where malicious instructions in project metadata could attempt to influence the reviewer agent's evaluation. \n
  • Ingestion points: Variables {WHAT_WAS_IMPLEMENTED}, {PLAN_OR_REQUIREMENTS}, and {DESCRIPTION} in code-reviewer.md. \n
  • Boundary markers: Absent. The templates do not use markers like triple backticks or XML tags to isolate untrusted content from the system prompt. \n
  • Capability inventory: The subagent has the ability to execute shell commands (git diff). \n
  • Sanitization: No validation or escaping of the interpolated variables is performed. \n- [COMMAND_EXECUTION]: The skill utilizes local shell commands (git rev-parse, git log, git diff) to identify and analyze code changes. These commands are used for their intended purpose in the code review process.
Audit Metadata
Risk Level
SAFE
Analyzed
May 7, 2026, 09:49 AM
Security Audit — agent-trust-hub — requesting-code-review