RTF Prompt Framework: Role, Task, Format

Study guide

RTF Prompt Framework: Role, Task, Format

Most people type a bare request into ChatGPT, Claude or Gemini and get a generic answer back. RTF fixes that with three short slots. Role tells the AI who to be. Task tells it what to do and gives it the material. Format tells it what shape the reply should take. Each slot is one line, and the whole prompt takes under thirty seconds to write. That speed is why RTF is the sensible default when you are unsure which framework to use.

This course is for busy, non-technical professionals who use AI for everyday work: drafting client emails, summarising meeting notes, comparing supplier quotes, writing Slack updates. You will see why a bare request leaves the AI guessing, how to write a role that changes the vocabulary and depth of the answer, how to open a task with one clear verb and paste the real material, and how to set a format you can paste straight into your file. You will also learn CTF, which swaps the role for the situation when the facts of your world matter more than expertise, and when to step up to CO-STAR, RISEN or RACE.

What is a prompt framework?

A prompt is the request you type into an AI tool such as ChatGPT, Claude or Gemini. A prompt framework is a reusable structure for that request: a short checklist of the parts a good prompt contains, usually named with an acronym so it is easy to remember. Why it helps: the AI fills in whatever you leave out with generic guesses. A framework makes sure you include the parts that change the answer, such as who it is for, what you want to achieve and the format you need.

Frameworks make prompts repeatable: you fill the same parts each time, save the prompt as a template and share it with colleagues. They are not magic words: the value comes from the detail you put in each part. They help most when a request is complex or open to interpretation; a quick, simple question can stay short. There are well over sixty named frameworks, and most of them remix the same ingredients: role, task, context, audience, format, constraints and examples.

In this course you learn RTF, which is a fill-in-the-slots framework: an acronym where each letter stands for one part of your prompt. It is best for: quick everyday requests; the default when you are in doubt.

Why three slots beat a bare request

A bare request leaves three gaps open: who the AI should be, who is reading the answer, and what shape the answer should take. The model fills open gaps with the most average version of everything, which is why the result feels generic. It is not that the AI is lazy. It has no information to work with, so it produces the middle of the road.

RTF stands for Role, Task, Format. Each slot is one line. Role sets the voice, Task sets the action and the material, Format sets the shape of the reply. Because the framework is so short, it takes under thirty seconds to write, which makes it practical for everyday work rather than a special occasion.

Vague slots produce vague answers. A role unrelated to the task gets ignored, and a format like be professional gives the model no shape to follow. Picture a project manager telling a client about a late delivery. A bare request returns a generic email full of filler. An RTF prompt returns a usable draft with the right tone, the key facts and a clear next step. RTF is a starting point, not a ceiling. Once the three slots feel natural, you can add context or examples.

Role: who the AI should be

The Role is the first slot in RTF and it works before the AI reads your task. It decides which words the AI reaches for, how deep it explains, and what it treats as important. A weak role produces a generic answer no matter how good the task is.

Specificity beats importance. A role with seniority, industry and a point of view, such as a senior recruiter in hospital settings, gives the AI far more to work with than act as an expert. It removes guesswork about level, audience and tone. Anthropic's guidance is that even a single sentence of role in the system prompt focuses behaviour and tone. You do not need a paragraph of backstory, just one sharp line that points the AI in the right direction.

Compare roles before you commit. Running the same task as an employment lawyer and as a friendly HR assistant shows how much the role changes the answer, and that comparison is the fastest way to learn what a role can and cannot do. A role is not a costume. It changes the priorities of the answer, so if you ask for a legal role you will get risk warnings even when you wanted plain reassurance. Choose the role that matches the answer you actually need.

Task: one clear verb and the material

The verb is the action word that opens your task. Draft, summarise, compare, rewrite and extract each produce a different kind of answer from the same material. Google's guide calls the verb the most important part of a prompt. If you leave it out, the AI picks its own action, and it often picks the wrong one.

Paste the material directly into the task. Notes, data, contract clauses and earlier versions all work. Context beats clever wording, because the AI can only work on text it can see. Keep one task per prompt. If you need three things done, send three steps. A prompt with three tasks inside it tends to produce three weak answers instead of one strong one.

A task with no verb and no material forces the AI to guess. It will invent a generic answer that looks plausible but does not use your content. Naming the verb and pasting the text removes the guess. You can add a scope line after the material, such as focus only on payment terms, and a limit such as keep it under 150 words. These narrow the answer without adding a second task.

Format: the shape of the answer

Format is the third slot in RTF and it controls three things at once: structure, length and tone. Structure is the layout, such as a table with named columns or three bullets. Length is how much text, such as 120 words. Tone is how it sounds, such as formal or friendly.

Without a format, the AI defaults to a generic essay that you then have to rewrite. With a format, the answer arrives in the shape you need, so it can be pasted straight into your work. The only change is one short line in the prompt. The fastest way to set format is to name the destination. Say where the text will live: an email body, a slide's speaker notes, a Slack message, or a spreadsheet row. The AI then writes for that place, at the right length and tone, without you spelling out every rule.

When you ask for a table, always name the columns in order, exactly as they appear in your file. If you leave the columns unnamed, the AI invents its own headings, and they rarely match your spreadsheet, so you spend time fixing them. Format works best when it is concrete. Prefer around 120 words over short, and warm and formal over nice. Concrete details give the AI no room to guess, which is what makes the output usable on the first try.

CTF: when context matters more than role

CTF stands for Context, Task, Format. It is the same three-slot habit as RTF, with the role replaced by the situation the AI cannot guess. Use CTF when the answer depends on facts about your world, such as a date, a place, a rule or a group of people. Use RTF when tone or expertise is what shapes the reply.

Context should be short. One or two lines of facts that would change the answer. Long background stories bury the task and weaken the reply. The two frameworks combine. Role plus context plus task plus format is still a prompt you can write in about thirty seconds.

A good test before you send is to ask what the AI cannot guess here. Whatever that is, put it in the prompt first. If the answer depends on your company's refund window, say so. If it depends on who is reading the message, say so. The AI cannot see your world, so anything it cannot guess has to come from you.

RTF in your job, and when to step up

A template is an RTF prompt you save and reuse. The Role, Task and Format stay fixed, and only the material in the middle changes each week. Three templates cover most weekly work: a status update, a summary of notes, and a first draft reply to a common request. Saving them turns RTF from a technique into a habit.

Step up to CO-STAR when audience and tone matter more than the role, because it adds context, audience, tone and response style. Step up to RISEN when the task has steps, because it carries role, instructions, steps, end goal and narrowing in one prompt. Step up to RACE when you need sources and caveats, because it asks the AI for evidence and honest limits rather than confident guesses.

When you are unsure what context the AI needs, add one line asking it to ask you up to five questions before it answers. That single line often saves a round of rewriting. RTF is the default for fast drafts. The bigger frameworks are for the jobs where audience, steps or sources change the answer.

Frequently asked questions