code-review

Pass

Audited by Gen Agent Trust Hub on Mar 16, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is purely instructional, providing guidelines, checklists, and examples for high-quality code reviews. It encourages developers to search for security vulnerabilities like SQL injection and XSS.
  • [PROMPT_INJECTION]: The skill describes a process for reviewing untrusted code, which is an inherent surface for indirect prompt injection.
  • Ingestion points: Source code from Pull Requests or local files processed during reviews.
  • Boundary markers: None specified in the skill body to separate instructions from review data.
  • Capability inventory: Mentions executing local test runners (e.g., npm test, pytest, cargo test) and using a task_edit tool.
  • Sanitization: Not explicitly defined in the provided review instructions.
  • [COMMAND_EXECUTION]: The skill references common development commands (e.g., npm test, pytest, cargo test) within an educational context for self-review. These are appropriate for the skill's stated purpose and do not represent a malicious attempt to execute arbitrary code.
  • [DATA_EXFILTRATION]: No patterns of unauthorized data access or exfiltration were detected. The skill uses generic placeholders in its code examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 16, 2026, 06:15 PM
Security Audit — agent-trust-hub — code-review