// Lesson 3 of 3 · ~25 min · Beginner

Prompts that work

After this lesson you can turn a vague request into a prompt that says what you want, why, in what shape, with an example.

You will be able to

Write explicit instructions instead of hoping the model guesses
Explain the why behind a rule so the model generalises it
Use examples, XML tags and layout to control a prompt’s structure and output format

Warm-up · Activity 1 of 7
Recall from the last lesson: what is a strong first step before answering a question about a long document?
Answer from general knowledge, then checkExtract word-for-word quotes relevant to the questionAsk for the shortest possible answer
Predict · Activity 2 of 7
Predict: which prompt is more likely to produce a rich, fully featured result?
B: “Create an analytics dashboard. Include as many relevant features and interactions as possible. Go beyond the basics to create a fully-featured implementation.”Both give the same result: the model knows what a good dashboard isA: “Create an analytics dashboard”
Practice · Activity 3 of 7
Why is “Your response will be read aloud by a text-to-speech engine, so never use ellipses since the engine will not know how to pronounce them” better than “NEVER use ellipses”?
Capital letters make a model ignore an instructionThe reason lets the model generalise, e.g. avoiding other things a speech engine cannot sayIt is not better, just more politeLonger prompts always work better
Practice · Activity 4 of 7
Match each tag to what you would put inside it.
<instructions> — Match for “<instructions>”
<context> — Match for “<context>”
<input> — Match for “<input>”
<example> — Match for “<example>”
Practice · Activity 5 of 7
Your app needs replies in an exact JSON shape, but the model keeps adding a friendly sentence before the JSON. Which change is most likely to fix it?
Ask the same question againAdd “please” to the promptMake the prompt as short as possibleAdd a few examples showing the exact output, and say to reply with the JSON only
Brain teaser · Activity 6 of 7
Brain teaser: apply the golden rule. A colleague with no background reads “Summarise these customer reviews and list the problems briefly.” What will they most likely need to ask first?
Are these reviews real?Nothing, the prompt is completeWhat font should the list be in?Who is the summary for, and how short is “briefly”?
Apply · Activity 7 of 7
Mini-task. Rewrite this prompt using what you practised: “Write something about our new feature for customers.”
Check your work against this list
- It states the task and the audience explicitly
- It says why, or what the text will be used for
- It separates the input material from the instructions, e.g. with tags
- It specifies the output format and length
- It includes or points to an example of the desired style

Exit ticket

5 questions, no hints. Score 80% or more to complete the lesson.

Finish every activity above to unlock the exit ticket.

Key ideas

Be clear and direct

Modern models respond well to clear, explicit instructions, and follow them precisely: ask “can you suggest some changes?” and you may get suggestions rather than the changes. If you want more than the basics, ask for it, and be specific about the format and constraints. Anthropic’s golden rule: show your prompt to a colleague who has minimal context on the task. If they would be confused, the model will be too.

Say why, show what

Explaining the reason behind an instruction helps the model apply it to cases you did not list. Examples are one of the most reliable ways to steer format, tone and structure. Make them relevant and varied, so the model does not copy an accidental pattern, and wrap them in <example> tags so they are not mistaken for instructions.

Separate the parts

Wrapping each type of content in its own tag, such as <instructions>, <context> and <input>, reduces misinterpretation. Tags can steer the output too: asking for the answer inside, say, <summary> tags is one of the formatting techniques the guide lists. With long documents, put the documents near the top and your question at the end. Anthropic reports that this can improve response quality by up to 30% in their tests, especially with complex, multi-document inputs.

Sources

Claude Docs — Prompting best practices

Last reviewed September 28, 2026

Prompts that work

Recall from the last lesson: what is a strong first step before answering a question about a long document?

Predict: which prompt is more likely to produce a rich, fully featured result?

Why is “Your response will be read aloud by a text-to-speech engine, so never use ellipses since the engine will not know how to pronounce them” better than “NEVER use ellipses”?

Match each tag to what you would put inside it.

Your app needs replies in an exact JSON shape, but the model keeps adding a friendly sentence before the JSON. Which change is most likely to fix it?

Brain teaser: apply the golden rule. A colleague with no background reads “Summarise these customer reviews and list the problems briefly.” What will they most likely need to ask first?

Mini-task. Rewrite this prompt using what you practised: “Write something about our new feature for customers.”