Prompt Framework Builder (RACE · CO-STAR · RTF · CRISPE) | RECATOOLS

How to Use the Prompt Framework Builder

Pick a framework

Choose the structure that fits your task — CO-STAR for content where tone and audience matter, RACE or RTF for quick everyday jobs, CRISPE for creative or exploratory work, TAG when you care most about the outcome.

Fill the slots

Each framework breaks the prompt into the parts an LLM actually needs — role, context, the exact task, and the output format. Fill what's relevant; leave a field blank and it's simply omitted.

Copy the assembled prompt

The tool builds a clean, sectioned prompt as you type. Copy it and paste into ChatGPT, Claude, Gemini, or any model. Nothing you type is sent anywhere — it all runs in your browser.

Iterate

Got a so-so answer? Change one slot — the audience, the format, the tone — and keep the rest. Tweaking one variable at a time is how you turn a decent prompt into a reliable one.

Prompt Frameworks: Structure Beats Guesswork

Why a framework beats a one-line prompt

Most people type a vague request — "write me a marketing email" — and get a vague, generic answer back. A prompt framework is simply a checklist that forces you to supply the things a large language model actually needs to do good work: who it should be, what exactly to do, the context around the task, and the shape of the output you want. The gap between that one-liner and a CO-STAR prompt that names the audience, sets the tone, states the objective, and asks for a five-bullet format is the gap between a throwaway draft and something you can ship.

Frameworks also make prompts reusable and testable. Because each part lives in its own slot, you can change one variable — swap the audience from "executives" to "beginners", or the format from prose to a table — while keeping everything else fixed, then compare the results. That is how serious prompt engineering works: not by re-typing a wall of text each time, but by treating the prompt as a small structured object you can tune. This builder keeps that structure for you and assembles a clean, sectioned prompt you can paste into any model.

"A model can only answer the question you actually asked. A framework's whole job is to make sure you asked a complete one."

Which framework, and when

They overlap on purpose — all of them push you to add a role, context, and an output format — but each has a centre of gravity. RACE (Role, Action, Context, Expectation) and RTF (Role, Task, Format) are fast and dependable for everyday tasks. CO-STAR adds Style, Tone, and Audience, which makes it the strongest choice for writing and analysis where voice matters. CRISPE leans creative and exploratory, with an "Experiment" slot that asks for alternatives. TAG (Task, Action, Goal) is the most compact and is outcome-focused. None of them are magic — the model still does the work — but a well-structured prompt removes the back-and-forth and gets you to a usable answer faster. Pick one, fill the slots, and let the structure do the heavy lifting.

10 Facts About Prompt Engineering

01

A prompt framework is a checklist of the parts an LLM needs: role, task, context, and output format.

02

CO-STAR stands for Context, Objective, Style, Tone, Audience, Response — it won a Singapore GovTech prompt competition.

03

Naming the output format ("a JSON object", "a 5-row table") is one of the highest-leverage things you can add.

04

Giving the model a role ("act as a senior editor") reliably sharpens tone and depth.

05

Specifying the audience changes everything — "explain to a 5-year-old" vs "to a CFO" yields very different answers.

06

RTF (Role · Task · Format) is the minimal framework — three slots and you're done.

07

Frameworks make prompts testable: change one slot, hold the rest, and compare outputs.

08

The same framework works across ChatGPT, Claude, Gemini and most models — it's vendor-neutral structure.

09

Long context first, instruction last often helps — but a clear structure matters more than ordering tricks.

10

This builder runs entirely in your browser — your prompt is never sent to any AI or server.

Frequently Asked Questions

A prompt framework is a reusable structure that splits a prompt into the parts a language model needs — typically a role, the task, the context, and the desired output format. Filling each part deliberately produces more complete, consistent answers than a single free-form sentence.
Use RACE or RTF for quick everyday tasks, CO-STAR when writing or analysis needs a specific tone and audience, CRISPE for creative or exploratory work, and TAG when you mostly care about the outcome. They overlap a lot — pick whichever slots match the thinking you'd do anyway.
Yes. The assembled prompt is plain, vendor-neutral text — copy it into ChatGPT, Claude, Gemini, Mistral, or any other model. Frameworks describe structure, not a specific provider's syntax.
No. The tool assembles your prompt entirely in your browser with plain JavaScript. Nothing you type is sent to any model, server, or third party, and nothing is stored.
No. Leave any slot blank and it's simply omitted from the assembled prompt. Fill the parts that matter for your task; even two or three well-chosen slots beat a vague one-liner.
Context, Objective, Style, Tone, Audience, and Response (format). It is a popular all-round framework because it explicitly separates tone and audience from the objective, which is exactly where generic prompts tend to fall down.
Models default to a safe, middling format when you don't ask. Saying "return a JSON object with these keys" or "a five-bullet summary" removes that guesswork, makes the output easier to use or parse, and reduces the need for a follow-up message.
The output box shows the assembled prompt live. The cleanest approach is to refine the individual fields, but you can also paste the result into your editor and adjust it further before sending it to a model.
No — the model still does the work, and results vary. A framework removes the most common cause of bad answers (an incomplete or ambiguous request) and makes your prompt easy to improve, but it is structure, not magic.
Completely free, with no account or sign-up. It runs in your browser and collects no data. Use it as many times as you like.

Related News

You may be interested in these recent stories from our newsroom.

No related news yet for this tool. Our editorial team publishes new pieces every week.

Browse all news →

Prompt Framework Builder