Some test cases require more than one user prompt to drive the agent to complete the task. For example, for deployment tasks, the agent may pause at any time to confirm resource types, locations and other parameters. This is a valid behavior. We also don't want to give the agent too much information upfront since this deviates from how an average user interacts with an LLM agent (aka. average users don't give full details of what they want to do upfront). Allowing the test author to pre-write follow up prompts will allow waza to test more scenarios.