Quick prompting tip for coding:

Vague prompts are bad because the LLM may fill in the gaps in ways you don't want.

INSTEAD: Feed the vague prompt to an LLM and ask it to output a prompt with all the vague parts filled in.

If you agree with how it filled in the details, use that instead of the original prompt. If you disagree with something, you can tweak it before using it.

Example prompt:

"I have a vague prompt for a frontend generator. Fill in the details to make it concrete. Use good taste: {PROMPT}"

Bonus tip: you can copy-paste this across chatgpt/grok/gemini to get different versions of the detailed prompt.