code-review
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs Git operations such as merge-base, diff, log, and show to evaluate code changes. It also executes a local Node.js script, check-identity.mjs, which uses execFileSync to securely interact with the Git binary.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data, including Git repository content and criteria specifications.\n
- Ingestion points: Repository commit data (git diff, git log) and criteria manifests (Markdown and JSON files) are read by the reviewer contexts.\n
- Boundary markers: The skill architecture mandates the use of two independent, fresh, read-only contexts for 'Standards' and 'Spec' reviews to ensure logical isolation between different review tasks.\n
- Capability inventory: The skill can execute Git commands and a local validation script, and it reads/writes local evidence artifacts in a specified evidence package folder.\n
- Sanitization: The check-identity.mjs script implements rigorous validation for JSON structure, OID formats using regular expressions, and SHA-256 digests to verify the integrity and provenance of all review artifacts.
Audit Metadata