Contain the incident without losing the evidence or the handoff

Edit incident-plan.json. ALL jobs need method, owner, integer start and authorityRef. End=start+immutable duration. A whole half-open [start,end) action needs a capable owner, exact actor/method authority (from<=start<end<=until), actor availability and capacity; shared overload blocks every overlapping job. Total planned cost must fit budget, or no job earns completion credit. Job and selected-method dependencies must finish with support by start; unsupported predecessors block dependents. Every capture must finish by its artifact deadline. Conservatively finish volatile capture before ANY planned shutdown starts, even if that shutdown lacks authority. SHA-256 consumes the exact hex fixture bytes. Custody[artifact] needs two ordered records {at,from,to,digest}: collector->custodian at supported capture end, then custodian->independent review owner at supported review end, with the same native digest. Messages for BOTH notice jobs need at, recipient, contained, verified, preserved and includesEvidence. At equals supported notice end; recipient is supplied; withhold evidence bytes. Boolean claims must match supported completed containment, verification and all captures as of at; null preserves uncertainty. The final notice requires all three true. Handoff acceptedAt equals supported handoff end, lead equals its owner, nextOwner and openActions preserve supplied accountability; custody/notice defects prevent a verified handoff. Report totalCost, containAt, outageMinutes=max(0,switchEnd-containEnd), harmUnits=containEnd*harmRate, readyAt=handoffEnd and evidenceDigests per artifact. Unsupported endpoints/digests use null. Meet harmLimit AND outageLimit; costs remain planned even when blocked. Repair a remaining custody, disclosure or report defect while preserving working jobs, without Reset.