requesting-code-review

Pass

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions and templates utilize shell commands, specifically git rev-parse, git log, and git diff, to programmatically retrieve commit identifiers and extract source code changes for the review process.\n- [PROMPT_INJECTION]: The code review subagent processes repository content (git diffs) and task metadata that are interpolated directly into the code-reviewer.md template. This architecture allows for indirect prompt injection if malicious instructions are embedded within the code being reviewed or the task descriptions.\n
  • Ingestion points: Git diff output and template variables ({DESCRIPTION}, {PLAN_OR_REQUIREMENTS}) in code-reviewer.md.\n
  • Boundary markers: The template does not utilize specific delimiters or isolation warnings to differentiate between system instructions and the untrusted data being analyzed.\n
  • Capability inventory: The subagent is authorized to read local repository files and execute git-related shell commands.\n
  • Sanitization: No input validation, escaping, or filtering is performed on the code content or descriptions before they are processed by the LLM.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 25, 2026, 12:58 PM