Prompt engineering frameworks are memory tricks — acronyms that make sure you never forget a part of a good prompt when it matters. Instead of staring at a blank chat box thinking “what should I include?”, you walk the letters: R… T… F… done. That’s the entire value: structure on demand.
There are dozens of frameworks floating around the internet — most are the same idea wearing different letters. Below are the 5 actually worth knowing, each with a filled example you can copy, plus a comparison table and a simple rule for choosing. They’re all built from the same anatomy we dissected in Anatomy of a Perfect Prompt — this is guide #7 of the Prompt Engineering roadmap.
The 5 Frameworks Worth Knowing
RTF — The Everyday Workhorse
R = Role · T = Task · F = Format
Best for: quick daily prompts — emails, summaries, explanations. Three parts, ten seconds to write, covers 70% of everyday use.
TCEF — When Style Matters
T = Task · C = Context · E = Example · F = Format
Best for: anything where the AI must match YOUR pattern — captions, product names, report sections. The E is few-shot prompting built into a formula.
CO-STAR — The Content Professional
C = Context · O = Objective · S = Style · T = Tone · A = Audience · R = Response format
Best for: marketing copy, blog posts, social content — anywhere voice and audience decide success. Award-winning framework from Singapore’s GPT-4 prompt competition.
RISEN — The Project Manager
R = Role · I = Instructions · S = Steps · E = End goal · N = Narrowing (constraints)
Best for: multi-step tasks and structured deliverables — plans, tutorials, processes. The S forces the AI to work in your sequence, not its own.
BAB — The Persuader
B = Before · A = After · B = Bridge
Best for: sales copy, pitches, cover letters — persuasion structures. You describe the problem state, the dream state, and ask AI to write the bridge.
Framework Comparison — Which One When?
How to Actually Use Frameworks (3 Rules)
- 1. Start with RTF for everything. Upgrade to a bigger framework only when the output disappoints — most tasks never need more than three parts.
- 2. Don’t fill parts with padding. If a letter has nothing real to say, skip it. “Style: good” is worse than no style line at all.
- 3. Save your filled frameworks. A CO-STAR you wrote for Instagram captions becomes a reusable template — swap the topic, keep the structure. That’s how a prompt library is born.
Build Your Own Framework (Yes, Really)
Once RTF feels automatic, customize. A student might use T-L-Q: Topic → my Level → Quiz me after. A blogger might use K-A-F: Keyword → Audience → Format. The letters don’t matter — what matters is that YOUR recurring task has a checklist you never skip. Test any framework you invent with the meta prompt: “Here’s my prompt structure for [task] — what am I missing?” More ready patterns in our 100+ examples.
Pick a framework. Fill the letters. Ship it. 🧩
60+ free guides on the full roadmap.
Frequently Asked Questions
A memorable acronym (like RTF or CO-STAR) that lists the parts a good prompt should include — a checklist that stops you from forgetting context, format, or constraints when writing prompts quickly.
RTF — Role, Task, Format. Three parts, ten seconds to apply, and it covers the majority of everyday prompts. Add more structure only when a task genuinely needs it.
Context, Objective, Style, Tone, Audience, Response format. It became famous after winning Singapore’s government GPT-4 prompt competition and is the go-to for marketing and content work.
Yes — frameworks are model-agnostic. They organize what YOU write; every major AI benefits from the same structure.
No — the labels are for your thinking, not the AI. Write the prompt as natural sentences; just make sure each part is present. (Labels don’t hurt, though, and can help in long prompts.)
A framework structures the prompt you write (RTF, CO-STAR). A technique changes how the AI processes it (chain-of-thought, few-shot). You combine them: a CO-STAR prompt can include ‘think step by step.’
