Ask HN: AI product builders, how do u assess/iterate on quality of your prompts?

I'm building an app that creates "choose your own adventure" stories. I'm looking to improve the prompt that writes these stories but I feel like I'm making changes somewhat blindly...I mean, I'm not sure if the changes I'm making are better for most users.

How do you assess / iterate on the prompts that power your apps? Any tips or tools that you'd recommend?