Verify that a person understands action-changing LLM or agent output before a dependent conversation, decision, handoff, forwarding, or consequential action continues. Use when correct understanding matters; skip routine, low-stakes exchanges unless the user explicitly requests a checkpoint.
