Organizational learning through trace analysis. Reads task execution traces (manifests, state, events, artifacts, handoffs) and produces structured retrospectives: what worked, what broke, structural observations, and proposed changes. Detects patterns across multiple runs — dead nodes, over-skipped roles, relations that never fire, critical path bottlenecks. Drives the Learning layer of the Graph+Policy+Ledger+Runtime+Learning model.
