What if you could stop your AI agent before it makes a mistake? · HackerLangs