I can't really provide reproducible prompts for my images because my workflows heavily use bouncing the latent between multiple different models. Sometimes with adjustments to prompts to direct the generation one way or another after the initial composition has been finalized.