If you're building an agent that talks to a real user, write the unhappy path first.
What does the agent say when the API fails?
When the user types nonsense?
When the model refuses?
Most agent demos fall apart on these three failure modes.
Designing them first will save you a week of bugs later.