How to write good prompts

A good prompt explains the task the way you'd explain it to someone new at your company: what you want, what it's for, what information you're starting from and what shape the result should take. You don't need a formula or a template; you need the AI not to have to guess. A prompt, by the way, is simply the message or instruction you write to it.

A few simple guidelines that work almost every time: give context (who you are, who's going to read the result), paste in an example of something you like, say what you don't want as well and, if the task is long, ask it for a plan first or for it to ask you questions before it starts. What confuses people most at the beginning is thinking a long prompt is better: what matters is that the information is useful, not how much of it there is. And the first attempt is rarely the right one; it's about talking, correcting and fine-tuning, just like with a person.

I didn't get this from a course: I've been learning it by using AI day to day, like anyone else. The principles that work for me with Claude and the other AIs, with before-and-after examples, I explain in "Asking AI for things without magic formulas". There I also explain why I don't think you need to pay for a "prompt engineering" course to get started.