Fig 1The agent loopPress play · ← → to stepNew task loaded at Plan
The task: one part, carried through every stationStations: Plan, Act, Observe, EvaluateTool rack: shell, web, fileThe agent loop as a rotary indexing table. Each quarter turn carries the task to the next station: Plan punches the plan into the part, Act fits a tool and adds real work, Observe gauges the result (tests passing, 0 to 6), and Evaluate either opens the gate to Done or sends the task round again. This run takes three loops: 4/6, 5/6, then 6/6.