The simple answer is that you cannot do so with the LLM itself, which does not follow any instructions at all but just predicts, one token at a time, what an an

The simple answer is that you cannot do so with the LLM itself, which does not follow any instructions at all but just predicts, one token at a time, what an answer that looks like following the instructions might contain. That's very different from actually following instructions! What you need is a "wrapper" application that audits and controls user input before submitting it to the LLM itself, and also examining and controlling the output before it is shown to the user. You've just discovered the real "moat" in building practical LLM apps that go beyond chat. You might like this article [0] that I wrote about how LLM models are built and why this is necessary. [0] https://medium.com/gitconnected/something-from-nothing-d755f49d6636 https://medium.com/gitconnected/something-from-nothing-d755f...