Isn't it both on LLMs? The input is your ability to craft a prompt, the output is checking if the prompt worked.