MMOOZZGG.com

AI literacy

Prompt builder: how to build a good request to an AI model

4 min read2 sources

The prompt builder puts together a request to an AI model from 7 parts: task, role, context, example, format, limits and answer language. Fill in the fields yourself or pick one of 6 templates, and the finished text appears in the Ready prompt window. You can copy it with one button and paste it into any AI model.

On this page · 4 sections
Templates

Who the AI should be

What to do and what result you need

Who you are, who the result is for, what matters

A text to follow

List, table, email, length

What not to do

Your text stays in your browser: the builder sends nothing to a server and stores nothing.

How to use the builder

A prompt is the text of an assignment for an AI model: what to do, for whom and in what form. The builder helps you not to forget any part.

1. Pick a template or start with an empty form. A template fills the fields with an example; replace the words in square brackets with your own.

2. Fill in at least the Task field: without it the prompt will not be built. The other fields make the answer more precise.

3. Choose the answer language: English or Russian.

4. Press Copy and paste the prompt into the AI model. The Clear button erases all fields.

Why each part matters and how to fix a weak answer is covered on the page about working well with an AI model.

What a good prompt is made of

Model developers give the same advice in their guides: a clear task, context, examples and a clear answer format. Anthropic compares the model to a smart new employee who does not know your rules: the more detail you give about the task, the more precise the result.

Task: what to do and what result you need. Role: who the AI model should be, for example an editor or a teacher. Context: who you are, who the result is for and what is important to know.

Example: a sample to follow. Google advises always adding examples to a prompt. Format: a list, a table or a letter, and how long. Limits: what not to do. Language: which language to answer in.

Weak: Write a letter to a client.

Strong: You write business letters. Tell the client that the delivery is 5 days late. The client expected the goods by 1 December. The letter is up to 10 sentences, with a subject line, and makes no promise of discounts.

Six templates

Letter: a business letter to a client or a colleague. The template asks you to say who you are, who the recipient is and what next step you need.

Post: a text for social media. You set the topic, the readers and the main idea, and in the Example field you paste a post whose tone you like.

Resume: a resume text for a job opening. The template forbids the AI model to invent experience and numbers and reminds you not to include your phone, address or passport details.

Document review: the main points, deadlines and obligations of the parties as a table. Paste the text without personal data or trade secrets.

Plan: steps to a goal with an owner and a deadline. Each step can be checked, so it is clear when it is done.

Topic explanation: in plain words, with an example from daily life and questions to test yourself. It suits a school student, an accountant and a manager.

What the builder does not do

The builder sends nothing to a server. The prompt is built right in your browser, and we do not see it.

The builder stores nothing. Close the page and the fields clear, so save a good prompt yourself.

The builder does not check your prompt. Check it yourself by the Anthropic rule: show the prompt to a person who does not know the task. If they get confused, the AI model will too.

Questions and answers

Does the builder save my prompts?

No. The text is built in your browser, nothing goes to a server, and the fields clear when you close the page. Copy the prompt you need and save it yourself.

Which AI model is the prompt for?

For any model that understands plain text: the prompt rules of different model developers are almost the same. Paste the prompt into a normal chat window, and choose the answer language, English or Russian, in the builder beforehand.

Sources and fact-checking

Checked by the editors on 16 October 2026.

  1. Anthropic, Prompting best practices, 2026 (platform.claude.com)
  2. Google AI for Developers, Prompt design strategies, 13 October 2026 (ai.google.dev)

Found a mistake? Tell us

Publications

Contact us