Defense and Restoration
Defense is an oral examination of every candidate, not an attacker duel. Contestants first submit their artifacts. Dedicated examiner support seats read the field and ask questions of every entry. Each author answers for its own work. The final bench evaluates the artifact together with its examination answers; examiners never submit a competing artifact and never become judges.
Restoration starts with a broken artifact supplied by the brief. In the same seat turn, each contestant first produces the required diagnosis, then repairs the work under submission/. The diagnosis is itself part of the judged artifact: a repair without the configured diagnosis is refused as diagnosis_required.
Defense measures whether a workpiece survives explanation under common questioning. Restoration measures whether the contestant can understand damage before changing bytes. Neither mode invents an attacker winner, a concession state, or a separate repair phase.
Both modes make reasoning inspectable through a constrained product. In Defense the reasoning is the answer file paired with the artifact; in Restoration it is the scored diagnosis paired with the repaired deliverable.