requesting-code-review

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill establishes a non-destructive review process by explicitly instructing the subagent to maintain a read-only state and avoid mutating the working tree. The use of git worktree in a temporary directory for inspection is a secure method for handling different code revisions without affecting the active environment.\n- [INDIRECT_PROMPT_INJECTION]: As an analysis tool that processes code and requirement documents, the skill possesses an inherent surface for indirect prompt injection. This is documented as a risk factor rather than a detected vulnerability.\n
  • Ingestion points: The subagent ingests code diffs from git diff and instructions from the [PLAN_OR_REQUIREMENTS] placeholder in code-reviewer.md.\n
  • Boundary markers: The template uses structural headers to organize input, but lacks formal delimiters (e.g., XML tags or code blocks with ignore instructions) for external content.\n
  • Capability inventory: The subagent is authorized to execute read-only Git commands and create temporary worktrees.\n
  • Sanitization: No sanitization or escaping is performed on the ingested code or plan text before it is presented to the subagent.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 01:43 PM
Security Audit — agent-trust-hub — requesting-code-review