TAG
Task · Action · Goal — The smallest useful structure. Three sentences that separate the subject, the operation, and the reason.
Use it for
Fast brainstorming, single questions, and moments when you want the model to have latitude. The Goal slot is the trick: telling the model why usually beats telling it how.
Avoid it for
Deliverables with a required shape. TAG says nothing about format, audience, or tone, so output length and structure drift between runs.
The slots
Each letter is a question the prompt has to answer.Task
What is the subject?
State the situation or object of work in one sentence. This is the noun, not the verb.
Example: Our team retrospective produced 40 raw comments about the last quarter.
Action
What operation should be performed on it?
The verb. Cluster, rank, critique, translate, expand, compress.
Example: Group them into themes and rank the themes by how often they recur.
Goal
What is this for?
The outcome you need downstream. This is the slot that earns TAG its keep — it lets the model resolve ambiguity in your favour.
Example: So I can pick the three problems worth putting on next quarter’s roadmap.
Strengths
- The Goal slot gives the model intent, which lets it make sensible judgement calls you did not anticipate.
- Almost no overhead — usable in a chat window without planning.
- Works well as a first pass before committing to a fuller framework.
Weaknesses
- Task and Action overlap in practice; people often write the same thing twice.
- No format, no audience, no constraints.
- Output is inconsistent across runs, which makes it a poor fit for anything repeatable.
Sources and grounding
- 1Prompting best practices — Add context to improve performanceAnthropic (Claude Platform docs)Anthropic documents that explaining why an instruction matters improves results — the mechanism behind TAG’s Goal slot.
- 2Prompt design strategiesGoogle (Gemini API docs)Google’s guidance on task-first instructions.