The line of code that lets your users rewrite your system prompt

Here is a line of code that ships in more LLM apps than anyone wants to admit: messages = [ {"role": "system", "content": f"You are a support agent. {user_input}"}, ] It runs. It answers question