How this scenario works
Represent connected components, state changes, constraints and feedback over time. For agent evaluation, the model focuses on agent policies, consistency, tool use and adaptation across repeatable situations.
Start with explicit assumptions, change one or more conditions, preserve state history, and compare why trajectories diverge.