The TAG Framework
Task, Action, Goal
TAG is a lightweight, outcome-first framework: name the Task, spell out the Action you want the model to take, and state the Goal it's serving. Because it ends on the goal, TAG nudges the AI to optimise for why you're asking — which often produces more useful answers than prompts that only describe the surface request.
Last updated · By the Prompt Orange team
Best for
Goal-driven requests where being clear about the outcome matters more than the persona.
What each part means
TAG stands for Task, Action, Goal. Here's what to put in each slot.
Task
The high-level thing you're working on — the subject the request sits inside.
Example: I'm preparing for a job interview for a product manager role.
Action
The specific action you want the AI to perform right now.
Example: Generate ten likely interview questions and a strong sample answer for each.
Goal
What success looks like — the outcome the output should help you reach. This steers the model's judgement.
Example: So I can practise and walk in feeling confident and well-prepared.
The TAG template
Copy this, fill in the brackets, and paste it into ChatGPT, Claude, Gemini, or any AI tool.
Task: [The broader thing you're working on]. Action: [The specific action you want the AI to take]. Goal: [The outcome you want this to achieve].
Before & after: TAG in action
See how the framework turns a vague prompt into a strong one.
“Give me some interview questions for a product manager job.”
Too vague—AI has to guess what you want
“Task: I'm preparing for a product manager interview at a mid-size fintech. Action: Generate ten likely interview questions covering product sense, prioritisation, and stakeholder management, with a concise model answer for each. Goal: So I can rehearse realistic scenarios and walk in confident.”
Specific, clear, ready to use
Why this works:
Stating the Goal changes the output. Without it, the model lists generic questions. With it, the model tailors difficulty, picks relevant themes (product sense, prioritisation), and frames answers as rehearsal material — because it knows what the questions are *for*.
Tips for getting the most from TAG
TAG shines when the outcome should shape the answer — interview prep, study plans, pitch practice, decision support.
Be honest in the Goal about constraints ("in 30 minutes", "for a non-technical audience") so the model can trade off accordingly.
If you also need a particular persona or output shape, combine TAG's Goal with RTF's Role and Format.
Keep the Action a single verb-led instruction; if it has multiple steps, consider RISEN instead.