The andon cord
Anyone on the line can stop the line, and stopping it is the job.
Taiichi Ohno, Toyota Production System: Beyond Large-Scale Production
1978, English edition 1988
On a Toyota line, a worker who sees a defect pulls a cord that signals a problem and, unresolved, halts production. The radical part was never the cord. It was that stopping the line was treated as correct behavior rather than as a cost, so problems surfaced immediately instead of traveling downstream inside finished work.
This is the direct ancestor of agent escalation design. The modern question is who pulls the cord when the process completes in eight seconds and nobody is watching, which is why the gate insists the trigger must fire by itself.
Which decisions may an agent make alone, and who is able to stop it?
What limits throughput, and what is supposed to happen when the line stops?
What would you show an auditor, and can the system produce it without being asked?
