What happens if you add the rules to the initial prompt and something like: "Think step by step, explain your move and if it is illegal rethink"

Very cool! Are you doing prompt engineering, fine-tuning, both, something else?

I'm wondering if it'd be cool to have a chess contest where all the bots are LLM powered. Seems to me like the contest would have to ban prompt engineering -- would have to have a fixed prompt -- otherwise people would sneak chess engines into their prompt generation.