The problem is not the LLM, the problem on the other side of the keyboard. No matter how good they are, LLMs can't read minds, they can only guess from what you write, and they won't be able to help you unless you give enough information to express the problem. And it is not an issue specific to LLMs, we already do "prompt engineering" when we are talking to humans. We don't call it that, but it is the same idea: write your message carefully to make sure the person on the other side understands your request.