Comment by tvmalsv

How well would it work to ask the working LLM to rewrite the prompt to get the best results? Do the models understand enough about themselves to do that?