Runs every generator live (section × difficulty) and applies the Writer → Candidate → Examiner check. Cell = % clean (5 unique options + Examiner-pass for composite). C = composite GL-grade share.
Green ≥ 99.5% · Amber 98–99.5% · Red < 98%. Each cell tests 2000 generated questions.
Mark each workstream item and add notes. Persists to build_status (course = 11+ NVR); read each session.
Eyeball real output. Correct answer ringed green.