Often times I come up with a prompt, then stick the prompt in an LLM to enhance / identify what I’ve left out, then finally actually execute the prompt.