Why Use Prompt Engineering Frameworks?
Large Language Models (LLMs) operate best when provided with clear boundaries, structures, and goals. Haphazard queries like *"write a blog post about cars"* result in generic and repetitive outputs. By utilizing a structured framework, you force the AI to assume a specific persona, consider relevant constraints, and follow strict output styles.
What is a Prompt Framework?
A prompt framework is a structured method for communicating with an AI. By providing specific elements (like Role, Context, and Goal) in a predictable pattern, you eliminate ambiguity and guide the AI to produce exactly the format, tone, and content you need. It turns random guessing into a repeatable science.
Framework Examples & Samples
APE (Action, Purpose, Expectation)
Best for simple, direct tasks where you just need something done quickly.
Action: Write a product description for our new wireless headphones
Purpose: To showcase key features and drive sales on our e-commerce website
Expectation: 3 compelling paragraphs with product benefits and a call-to-action
TAG (Task, Action, Goal)
Highly goal-focused framework for when the end metric is what matters most.
Task: Create a weekly social media posting schedule
Action: Draft 5 posts targeting marketing managers
Goal: Increase click-through rate to our blog page by 15%
RACE (Role, Action, Context, Expectation)
Excellent for business writing where the AI needs a specific persona.
Role: Expert Copywriter
Action: Write an email newsletter pitching our new product
Context: Our audience consists of busy startup founders who value efficiency.
Expectation: Short, punchy, professional tone, with a clear call-to-action button text
CARE (Context, Action, Result, Example)
Context-rich format when you have a specific example or reference style.
Context: We are launching a new online community for software developers
Action: Write a welcoming email sequence for new members
Result: Get new members to introduce themselves and complete their profiles
Example: e.g., 'Welcome to DevCircle! We're glad to have you...'
RISE (Role, Input, Steps, Expectation)
Process builder for complex workflows that require sequential steps.
Role: Financial advisor
Input: Income data, savings goals, and investment preferences
Steps: 1. Analyze cash flow, 2. Recommend asset allocation, 3. Create tax plan
Expectation: A structured PDF-ready summary with headers and bullet points
ERA (Expectation, Role, Action)
Results-first framework for when the output format is the most critical constraint.
Expectation: A comparative table comparing features of Next.js vs Remix
Role: Senior Web Developer
Action: Audit both frameworks focusing on performance and developer experience
CREATE (Character, Request, Examples, Adjustment, Type, Extra info)
Advanced structure for creative projects and complex technical tasks.
Character: Veteran Tech Lead
Request: Perform a code review of a React state management component
Examples: Include code snippets showing good vs bad practices
Adjustment: Make comments constructive, educational, and friendly
Type: Markdown table listing issues, files, and refactored code blocks
Extra info: Focus heavily on minimizing re-renders and memory leaks
TRACE (Task, Request, Action, Context, Expectation)
Perfect for detailed project outlines and professional deliverables.
Task: Prepare a client proposal document
Request: Outline the project timeline and milestones
Action: Draft estimated hours, deliverables, and cost breakdowns
Context: Client is a mid-sized healthcare company upgrading their legacy database.
Expectation: A clean proposal draft in markdown format
ROSES (Role, Objective, Scenario, Expected Result, Size/Style)
Used heavily in UX/UI design and specific scenario-based problem solving.
Role: UX Researcher
Objective: Design a user testing plan for a mobile banking checkout page
Scenario: Users are reporting checkout friction when adding a new credit card.
Expected Result: 5 usability test scenarios with questions and metrics
Size/Style: Professional tone, under 1000 words
SPARK (Situation, Problem, Action, Result, K-Factor)
Excellent for strategic problem solving and defining strict constraints.
Situation: We run a SaaS business with 5,000 monthly active users
Problem: Churn rate has increased by 4% over the last quarter
Action: Design a customer retention campaign focusing on inactive users
Result: A step-by-step campaign layout with email templates
K-Factor: Focus must be on automation - no manual tasks for support agents