The anatomy of a good prompt, and three myths to drop

A good prompt is not a secret phrase. It is just a clear request with the right pieces attached: a goal, some context, the output format you want, a few constraints, and maybe an example. Stack those five parts and you will get far better answers than any clever wording.

The nice thing is that none of this is technical. If you can write a clear note to a capable colleague, you can write a good prompt. Let us pull one apart.

What are the five parts of a good prompt?

Think of a prompt as a small stack. Each layer adds clarity.

1. The goal

Say what you want in one plain sentence. “Summarise this report for my manager in five bullet points.” That is the spine of the whole prompt. If you are vague here, no amount of polish later will save it.

2. The context

This is the part people skip, and it is the most important. Paste the actual report. Include the messy notes, the email thread, the audience, the deadline. The model cannot read your mind, so the more real material you hand it, the better the answer.

3. The output format

Tell it the shape: a table, three bullets, a short paragraph, a 280-character post. If you do not say, you get the default and then waste a turn reshaping it.

4. The constraints

These are your guardrails. “Keep it under 100 words.” “No jargon.” “British spelling.” “Friendly but not gushing.” Constraints are how you steer the tone and length without a back-and-forth.

5. An example or two

If you have a sample you like, show it. One or two examples teach the model your style faster than any description. This is the single biggest quality jump for tricky requests.

Which three myths should you drop?

A few habits hang around from the early days. Time to let them go.

Myth one: longer prompts are better. Length is not the goal, clarity is. A tight five-line prompt with real context beats a rambling page of instructions. Padding just buries the signal.

Myth two: there is one perfect prompt. There is not. Good results come from a first attempt plus a quick refine. Treating prompting as a one-shot puzzle sets you up to be frustrated. I unpack that mindset in prompting is editing, not incantation.

Myth three: you have to sound technical. You do not need keywords, role-play, or fake authority. Plain language works. In fact, the old “you are a world-class expert” routine adds almost nothing on capable models, which I covered in stop role-playing your prompts.

Put it together

Next time you open a chatbot, write your goal, paste the real context, name the format you want, add a couple of constraints, and drop in an example if you have one. That is the whole anatomy. No incantations, no costume, just a clear request the model can actually act on.