requesting-code-review

Warn

Audited by Gen Agent Trust Hub on Apr 11, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The template in code-reviewer.md constructs shell commands (e.g., git diff {BASE_SHA}..{HEAD_SHA}) using string interpolation. This creates a risk of command injection if the variables are populated with data from untrusted sources—such as a pull request description or externally suggested commit SHAs—that contain shell metacharacters.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its handling of untrusted data.
  • Ingestion points: External data is ingested through placeholders {WHAT_WAS_IMPLEMENTED}, {PLAN_OR_REQUIREMENTS}, and {DESCRIPTION} in code-reviewer.md.
  • Boundary markers: No delimiters (e.g., XML tags, triple quotes) or 'ignore embedded instructions' warnings are used to isolate these inputs from the agent's core instructions.
  • Capability inventory: The subagent is authorized to execute shell commands (git) and access the filesystem, providing a high-impact target for successful injection.
  • Sanitization: There is no evidence of sanitization, escaping, or validation of placeholder content before it is interpolated into the prompt.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 11, 2026, 06:19 PM
Security Audit — agent-trust-hub — requesting-code-review