review

Pass

Audited by Gen Agent Trust Hub on Aug 10, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill reviews untrusted documentation and source code from the workspace, which presents a surface for indirect prompt injection.
  • Ingestion points: Project source files, README, PRD, and jtbd.json files.
  • Boundary markers: The orchestration layer does not define specific delimiters for separating processed data from instructions.
  • Capability inventory: The agent can read local files, launch sub-agents, and perform browser-based walkthroughs.
  • Sanitization: Data is ingested for review purposes without additional sanitization logic.
  • [COMMAND_EXECUTION]: The skill can execute local project checks and the 'tokens contrast' utility to verify audit results, which is expected behavior for its design auditing purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 10, 2026, 09:33 PM
Security Audit — agent-trust-hub — review