stop-hook-needs-agent-stop-intent-detection-not-just-event-name
The current stop hook logic evaluates completion criteria based solely on the hook_event_name=‘Stop’ signal, which fires even when the agent pauses to await user input rather than genuinely terminating. The hook should detect whether the agent’s last message was a termination statement versus a clarifying question or holding response before running the verification checklist. A guard condition checking for actual stop() invocation or terminal language patterns would eliminate false positives.