The TAG Framework
Task, Action, Goal — three short pieces that turn a quick request into a clear one. TAG is the simplest of the five frameworks, built for speed: the one to reach for when you need a fast, useful answer and heavy context would just slow you down.
What is the TAG framework?
The TAG framework is the shortest checklist for writing a good prompt. Instead of dumping a half-formed thought into the chat box, you give the AI three things it almost always needs: the Task — what you are working with — the Action you want it to take, and the Goal, or the reason behind the request. That is the whole framework. It fits on a sticky note, and most of the time it is all the structure a quick request requires.
Large language models like ChatGPT, Claude, and Gemini are not mind readers. They answer by predicting what should come next from what you gave them, so a one-word request like "summarize" leaves them guessing at the length, the audience, and the point. TAG closes the most important gaps without making you write a paragraph of setup. Each of the three pieces removes one assumption the model would otherwise have to make on your behalf. (For the bigger picture on why this matters, see why better prompts matter.)
TAG is a natural fit for students and educators, because so much of the work is small and repetitive: rephrase this sentence, shorten this paragraph, draft three questions, simplify this definition. Those jobs do not need a persona or a strict template — they need to be clear and finished in seconds. TAG is the entry point to prompting: learn it first, use it constantly, and graduate to a fuller framework only when a task genuinely calls for one. The rest of this guide breaks down each element and shows TAG at work.
The three elements of TAG
Each card shows the question the element answers, with quick examples for studying and teaching.
Task (T)
What are you working with, in one line?
- •Here is a sentence from my lab report.
- •Here is a paragraph from the textbook chapter.
- •Here is the definition of "photosynthesis" from my notes.
Action (A)
What should the AI actually do with it?
- •Rewrite it in a more formal, academic tone.
- •Summarize it into a single sentence.
- •Write three discussion questions about it.
Goal (G)
Why do you need it — what is it for?
- •So it fits the tone of my final paper.
- •So I can put it on a flashcard.
- •So my 7th-graders can answer them in groups.
When to graduate from TAG
TAG covers the small stuff, but it deliberately leaves things out. There is no slot for a persona, a strict output format, or a multi-step procedure — and that is the point. The moment you find yourself wishing for one of those slots, the task has outgrown TAG. Want the AI to hold a role and explain at your level, or to return a long, well-organized document? Move up to CRAFT. Want it to follow a defined sequence of steps? Reach for RISE. Knowing when not to use TAG is part of using it well.
Why TAG works (even though it's simple)
TAG looks almost too short to make a difference, but each of its three pieces fixes a real failure mode. The Task gives the model the actual material to work on instead of a description of it — pasting the sentence beats saying "I have a sentence." The Action replaces a fuzzy verb with a specific one: "rewrite this more formally" is answerable in a way that "fix this" never is. And the Goal tells the model what "good" means here, so it can make sensible calls about length and tone without you spelling them out.
That third piece is the one people skip, and it is quietly the most useful. The Goal is the difference between a summary written for a five-year-old and one written for a colleague — same Task, same Action, very different result. When you tell the model the summary is "for a flashcard" or the questions are "for a group of seventh-graders," it adjusts the level and the framing automatically. You are giving it just enough to stop guessing, and no more.
You do not have to label the parts with the literal words "Task" and "Action" — the model does not need them — but running through the three in your head keeps you from firing off a half-formed request. That is the real value of such a tiny framework: it is a thinking aid that costs you almost nothing. And if you would rather not assemble even three pieces by hand, the TAG prompt generator gives you a field for each one and writes the finished prompt for you.
TAG in action: before and after
Two quick, everyday jobs — short on purpose. Notice how little it takes to remove the guesswork.
Make this sound better: "The experiment didn't really work out the way we thought."
[Action] Rewrite it in a more formal, academic tone, keeping it to one sentence.
[Goal] So it matches the professional voice of the rest of my report.
"Make this sound better" leaves the model guessing at how formal to go and how long to make it. The TAG version pins down all three in one short prompt — a single formal sentence that fits the report — and you get a usable rewrite on the first try.
Give me some discussion questions about this chapter.
[Action] Write three open-ended discussion questions about it.
[Goal] So my 7th-graders can talk them through in small groups for ten minutes.
Because the prompt names the grade level and the purpose — small-group talk, not a written quiz — the AI returns three open-ended questions pitched for twelve-year-olds, ready to drop straight into the lesson. No persona, no template, no fuss: that is TAG doing exactly enough.
Common mistakes to avoid
- Skipping the Goal. It is the easiest piece to drop and the one that does the most work. "Summarize this" with no reason gives you a generic summary; "so I can put it on a flashcard" gives you something you can actually use.
- Describing your material instead of pasting it. The Task should include the actual sentence, paragraph, or list. The model can rewrite text you give it; it cannot rewrite text it has never seen.
- Using a weak Action verb. "Fix this" or "help with this" tells the model almost nothing. Swap in a precise verb — rewrite, summarize, list, simplify, rephrase — so it knows the exact operation you want.
- Stacking several jobs into one prompt. TAG handles one clear task at a time. If you need a summary and questions and a rewrite, ask in three short turns — the model keeps the material between them.
- Forcing TAG onto a job that needs more. The biggest mistake is using TAG when the task really calls for CRAFT or RISE. If three slots leave you cramming, that is not a TAG prompt — it is a fuller framework waiting to happen.
When to use TAG (and when to upgrade)
TAG is the right tool when the request is small and there is only one moving piece — a quick rewrite, a one-line summary, a handful of questions. The moment a task grows past three slots, one of the sister frameworks will fit better:
- →CRAFT (Context, Role, Action, Format, Task) — for substantial, well-structured output like study guides, essays, and lesson plans.
- →RISE (Role, Input, Steps, Expectation) — for multi-step assignments where you want the AI to follow a procedure.
- →CO-STAR — when the audience and tone of the message matter most, such as emails or announcements.
- →CARE — when you need the output in a strict, repeatable format every time.
TAG is not a beginner's framework you outgrow — even when you are fluent in all five, it stays the right answer for everyday one-shot requests. Not sure which to pick? How it works walks through the whole methodology, and the use cases page shows the frameworks applied to real situations.
Frequently asked questions
What does TAG stand for in prompt engineering?
TAG stands for Task, Action, Goal. It is the simplest of the structured prompt frameworks: you state what you are working with, what you want the AI to do with it, and why — the purpose behind the request. Three short pieces are usually enough to turn a vague one-liner into a clear, answerable prompt.
Is TAG good for students and teachers?
Yes. TAG is built for the quick, everyday requests that fill a student's or teacher's day — rephrasing a sentence, summarizing a paragraph, drafting a few discussion questions, simplifying a definition. It gives the AI just enough direction to be useful without the overhead of a longer framework, so you get a usable answer in seconds.
How is TAG different from CRAFT?
TAG has three slots; CRAFT has five (Context, Role, Action, Format, Task). TAG is meant for small, one-off jobs where heavy context would be wasted effort. CRAFT adds an explicit Role and Format, which pay off when the output is substantial — a study guide, an essay, or a lesson plan — and the structure of the answer really matters.
When should I move from TAG to CRAFT or RISE?
Switch the moment three slots stop feeling like enough. If you catch yourself wanting to specify a persona, a strict output format, or a multi-step procedure, that is the signal that the task has outgrown TAG. Reach for CRAFT when you need a longer, well-structured piece, and RISE when you want the AI to follow a defined sequence of steps.
Which AI models does TAG work with?
All of them. TAG is model-agnostic, so the same three-part prompt improves results in ChatGPT, Claude, Gemini, Microsoft Copilot, and open-source models. Every large language model responds better when you state what you have, what you want done, and why — TAG just makes sure you say all three.
Ready to try it out?
Open the builder, fill in the three fields, and copy your finished prompt into ChatGPT, Claude, or Gemini.