Keep the attack engine. Add the context and evidence plane.

This capability plane preserves the offensive platform as the engine for generating, chaining, reproducing, and retesting attacks, while adding a AI Security LLC context and evidence plane that qualifies what is observed, what is inferred, and what remains uncertain. The new plane grounds attack traces in gray-box system context, identity, permission, and authority boundaries, then separates evidence-backed path segments from inferred ones so findings can be challenged and validated independently. It also introduces remediation chokepoints that cut across chains, prioritize fixes by leverage and reach, and return structured results to the platform with explicit grounding, residual uncertainty, and retest state. The result is a closed-loop system that keeps attack generation intact while turning raw attack output into evidence-qualified paths, actionable remediation priority, and structured platform return.