stop-hook-fires-on-agent-waiting-state-not-just-stop

The Stop hook fires repeatedly even when the agent sends a short acknowledgment message and is waiting for user input — not only when the agent actually terminates. This creates a false-positive loop where the hook evaluator sees a ‘Stop’ event but no actual stop() call or agent-initiated termination exists. The hook condition text (‘Before allowing the agent to stop’) is only meaningful when a real stop action is initiated; firing it on idle-wait messages floods the conversation and blocks nothing meaningful.