requesting-code-review

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data by interpolating git diffs and implementation descriptions into the subagent's prompt template.
  • Ingestion points: The variables {WHAT_WAS_IMPLEMENTED}, {PLAN_OR_REQUIREMENTS}, and {DESCRIPTION} in code-reviewer.md, along with the output of git diff commands.
  • Boundary markers: The prompt template lacks specific delimiters (such as XML tags or dedicated boundary tokens) to isolate untrusted data from instruction text.
  • Capability inventory: The skill performs read-only local git operations; no network or file-write capabilities were detected.
  • Sanitization: Input variables are interpolated directly into the markdown template without escaping or validation logic.
  • [COMMAND_EXECUTION]: The skill provides instructions for executing shell-based git commands (git rev-parse, git log, git diff) to identify changes. These commands are standard for the tool's stated purpose and are restricted to local repository metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 09:30 AM
Security Audit — agent-trust-hub — requesting-code-review