Scores AgentUI schema rendering and form behavior from real evidence.
Built-in suites describe real checks, but no pass/fail result is claimed until a run persists evidence.