Frontier Model Red-Team Report Writer

Turns raw red-team exercise artifacts (transcripts, capability elicitations, jailbreak traces, evaluator scores) into a structured red-team report suitable for a model card appendix, an SB 53 catastrophic-risk assessment section, and an EU AI Act GPAI systemic-risk disclosure — normalizes threat categories (CBRN uplift, cyber offense, autonomous replication, deception, loss-of-control), scores severity, quantifies uplift over baselines, and produces the write-up plus a residual-risk statement