CO-STAR is powerful, but it's overkill for a quick Slack reply or a one-line caption tweak. This lesson introduces RTF — Role, Task, Format — a lightweight three-field framework for the 80% of prompts that don't need the full six-variable treatment. Knowing when to use the heavy framework versus the light one is itself a skill, and it's what separates someone who "knows prompting frameworks" from someone who actually applies them efficiently all day.
What RTF Solves
Most everyday AI use isn't a client deliverable — it's drafting a quick reply, summarizing a page, rewriting a sentence. Using full CO-STAR for these tasks wastes time typing out Style, Tone, and Audience fields that don't materially change the outcome. RTF strips prompting down to the three fields that almost always matter:
| Initial | Component | Description |
|---|---|---|
| R | Role | Who the model should act as, briefly. |
| T | Task | The specific, single action to perform. |
| F | Format | The shape of the output you want back. |
RTF in Practice
Role: You are a helpful editor.
Task: Tighten this paragraph to half its length without losing the main point.
Format: Plain text, single paragraph, no bullet points.
Compare the time cost: writing that takes maybe 15 seconds, versus 45+ seconds for a full CO-STAR block. For a task you'll repeat dozens of times a day, that difference compounds fast.
When RTF Is Not Enough
RTF breaks down exactly where CO-STAR's extra fields earn their keep:
- The output is subjective or brand-sensitive (marketing copy, client-facing writing) → you need Style and Tone.
- Different readers would want different things (a report for a CEO vs. a technical team) → you need Audience.
- The task involves reasoning over a specific document or dataset → you need a dedicated Context field, not just an implied one.
A useful rule of thumb: if getting the tone or audience wrong would embarrass you in front of a client, use CO-STAR. If it would just mean a quick re-ask, use RTF.
Combining Both Frameworks in One Session
Experienced prompt engineers don't pick one framework and use it exclusively — they context-switch constantly:
9:03 AM — RTF: "Role: proofreader. Task: fix grammar in this WhatsApp
draft. Format: same length, plain text."
9:15 AM — CO-STAR: full client pitch email, all six fields specified.
9:40 AM — RTF: "Role: analyst. Task: summarize this article in 3 bullets.
Format: bullet list."
The framework is a tool you reach for based on stakes, not a religion you follow for every single prompt.
🇵🇰 Pakistan Angle
Time genuinely is money for Pakistani freelancers juggling multiple Upwork contracts at once — every extra 30 seconds spent over-engineering a low-stakes prompt is 30 seconds not spent on billable work or landing the next contract. Learning to downgrade to RTF for internal, low-risk, or repetitive tasks (drafting a quick client check-in message, summarizing a long brief before you read it fully) frees up the mental energy and time budget to use full CO-STAR where it actually earns you money — the client-facing proposal, the portfolio piece, the pitch that has to land.
Do This Now
Go through your last ten AI prompts from today or yesterday (check your chat history). For each one, decide: should this have been RTF or CO-STAR? Rewrite the three that were most mismatched — either an RTF prompt that needed CO-STAR's precision, or a bloated CO-STAR prompt that RTF would have handled in a fraction of the time. Notice how quickly you can now categorize a task's stakes before you even start typing.
Key takeaway: Not every prompt deserves six fields. RTF handles the fast, low-stakes majority of your AI usage; CO-STAR earns its extra structure only when tone, audience, or brand consistency are actually on the line.