Split train and tool simulator modules; mastery curriculum and grader workflow nudge.

Cleanup: drop legacy support-ops cruft now that DriftShield is the only collection

DriftShield Phase A: D1 collection, 4 runtime-failure tasks, ops.get_recommendation, recoverable schema drift

feat: GRPO training stack (Qwen3 + Unsloth opt-in, curriculum, eval, audit)

add expert task: GDPR compliance trap with misleading signals

clean rewrite: shorter vars, no comments, same logic