stop-hook-needs-agent-stop-intent-guard

The stop hook script must detect whether the agent actually called stop() versus merely pausing for user input. A guard condition checking for explicit stop intent (e.g., presence of a stop tool call in the transcript or a sentinel phrase) would prevent the false-positive loop. Without this guard, any waiting state in a multi-turn skill like superpowers:brainstorming triggers repeated hook evaluations for every idle cycle.