Prompt Frameworks
CRISPE, RTF, CRAFT
The Problem: With so many prompting techniques, how do you know which structure to use? Are there proven templates you can start with?
The Solution: Use Tested Recipes
Prompt Frameworks are structured templates and patterns that provide a starting point for effective prompts. They're like recipes for chefs — proven combinations that work, which you can customize to taste. Many frameworks incorporate prompt structure principles and can be enhanced with Chain-of-Thought or few-shot examples.
Think of it like tested recipes for chefs:
- 1. RISEN: Role, Instructions, Steps, End goal, Narrowing
- 2. CRISPE: Capacity, Role, Insight, Statement, Personality, Experiment
- 3. CO-STAR: Context, Objective, Style, Tone, Audience, Response
- 4. RTF: Role, Task, Format
Where Is This Used?
- Content Creation: Consistent, high-quality outputs
- Team Standards: Shared prompting conventions
- Learning: Teaching prompt engineering fundamentals
- Production Systems: Reliable, repeatable prompt structures
Fun Fact: The most important elements across all frameworks are typically Role, Context, and Task. If you include these three clearly, you're already doing better than 80% of prompts!
Try It Yourself!
Compare different frameworks and see how they structure prompts for better results.
📝 Frameworks help structure your prompt and remember everything. Choose a framework and fill in the fields — get a ready-made quality prompt!
Comprehensive structure for detailed prompts
Frameworks are a checklist for your prompt. They help you remember important details: who responds (role), what's needed (task), how to present (format/tone). Result: fewer revisions, more accurate answers!
| Framework | Components | Best For |
|---|---|---|
| RTF | Role, Task, Format | Quick technical tasks |
| RACE | Role, Action, Context, Expected output | Simple tasks with expected format |
| CRAFT | Context, Role, Action, Format, Tone | Universal balance |
| CRISPE | Capacity, Insight, Statement, Personality, Experiment | Creative tasks, copywriting |
| RISEN | Role, Instructions, Steps, End goal, Narrowing | Analysis, multi-step tasks |
| CO-STAR | Context, Objective, Style, Tone, Audience, Response | Marketing, audience-aware content |
Ready-made examples for each framework — copy and adapt to your task.
[Context] We are launching a meditation app targeting busy professionals. [Objective] Write 3 LinkedIn posts to attract early adopters for the beta test. [Style] Storytelling with personal examples. Like a TED talk, not an ad. [Tone] Warm and supportive, but data-driven. No esoterics — only science and practice. [Audience] Managers and entrepreneurs aged 30–45 experiencing chronic stress. [Response] 3 ready-to-post entries of 150 words. Each with a hook, story, CTA, and hashtags.
CO-STAR: Context → Objective → Style → Tone → Audience → Response
Try it yourself
Interactive demo of this technique
Write an email to a customer apologizing for a delivery delay
Dear Customer,
We apologize for the delivery delay. We are working on resolving the issue. Your order will be delivered soon.
Best regards, Support Team
Subject: Update on your order #A-4821 — new delivery date
Hi Maria,
I am reaching out personally to update you on your laptop delivery (order #A-4821). Due to a delay at our distribution center, delivery has been moved from March 15 to March 18. I understand how frustrating this is, and I sincerely apologize.
To make it right, we have:
- Upgraded your order to priority shipping (at no cost)
- Added 500 bonus points to your account
- Assigned you a dedicated account manager (that is me)
You will receive an SMS with the exact delivery window on March 18. If you have any questions, I am available via this email or at +1 (800) 555-0199.
Best regards, Anna Taylor Customer Success Manager
The CRAFT framework transforms a vague request into a precise brief. Each parameter (context, role, action, format, target) removes one degree of freedom, leaving the model less room to go off track.
Create a free account to solve challenges
3 AI-verified challenges for this lesson
This lesson is part of a structured LLM course.
My Learning Path