RGE Monitor Implements Ontological Trust for Long-Horizon Agent Safety

The RGE monitor tracks whether long-horizon agent trajectories drift from user-authorized objectives by decomposing trust into Role, Goal, and Evidence. It uses LLMs to generate structured representations while performing real-time trust-state updates and projections.