Prompt Framework
📚 Table of Contents
- 📚 Table of Contents
- 🧠✨ Introduction
- 🧩 5W1H
- ✅ Good Prompt vs ❌ Bad Prompt
- 🔗 Combination
- Final Thoughts 🚀
🧠✨ Introduction
🧠 Prompting is quickly becoming a core skill for developers, designers, and knowledge workers.
⏳ A good prompt can save hours of work, while a vague one can lead to confusing or unusable results.
🧩 Think of prompts as interfaces for thinking — the clearer your interface, the better the output.
🚀 This post introduces a simple, reusable Prompt Framework that helps you to:
- 🗣️ communicate intent clearly
- 🎯 get more consistent results
- 🔁 reuse prompts across tools and projects
🧩 5W1H
🧠 One of the easiest ways to structure a strong prompt is by borrowing the classic 5W1H framework.
👤 Who
👤 Who is the AI supposed to be?
- A senior engineer
- A product manager
- A teacher explaining to beginners
“Act as a senior backend engineer…”
📌 What
📌 What exactly do you want?
- Generate code
- Review text
- Explain a concept
“Explain how Kafka consumer groups work…”
🎯 Why
🎯 Why are you asking?
- Learning
- Production use
- Decision making
🧠 This helps the model choose the right depth and tone.
“…so I can explain it to junior developers.”
⏱️ When (optional)
⏱️ When are time or context constraints relevant?
- Current trends
- Backward compatibility
- Version-specific behavior
“Using Kafka 3.x…”
🌍 Where (optional)
🌍 Where will this be used?
- Web
- Mobile
- Backend
- React, Spring Boot, FastAPI
“…in a Spring Boot application.”
⚙️ How
⚙️ How should the output look?
- Bullet points
- Code only
- Step-by-step explanation
“Use simple examples and diagrams in text.”
✅ Good Prompt vs ❌ Bad Prompt
❌ Bad Prompt
“Explain Kafka”
🚨 Problems:
- 👤 No role
- 🎯 No goal
- 🌍 No context
- ⚙️ No format
- ❌ No example
🤔 The AI must guess everything, which often leads to generic or unfocused answers.
✅ Good Prompt
“Act as a senior backend engineer.
Explain Kafka consumer groups in simple terms for junior developers, using Kafka 3.x and a Spring Boot context.
Use bullet points and a short example, similar to explaining how HTTP load balancing works.”
✅ Why this works:
- 👤 Clear role
- 🎯 Clear goal
- 🌍 Clear context
- 📐 Clear format
- 🧩 Clear example
🧠 Examples act as anchors — they show the AI what “good” looks like, not just what to do.
🔗 Combination
🔗 The real power comes from combining frameworks.
🧠 Prompt Formula
Role + Task + Context + Constraints + Output Format + (Optional) Example
🔁 Reusable Prompt Template
Act as [ROLE].
Your task is to [WHAT].
This is for [WHY].
Context: [WHERE / WHEN].
Output requirements:
- Format: [FORMAT]
- Style: [STYLE]
- Length: [LENGTH]
- Example (optional): [REFERENCE OR SAMPLE]
🔁 You can store these templates and reuse them across:
- ChatGPT
- Claude
- GitHub Copilot
- Internal AI tools
4️⃣ Why This Works
🧠 This framework works because it separates structure from intent.
-
🧩 Who + How → quality control (template-owned)
🎛️ These define how the AI should think and respond.
📈 They improve output quality regardless of topic. -
🎯 What + Why → intent (user-owned)
🧠 These define what matters.
🙋 Only the user knows the real task and success criteria. -
🌍 Where / When → relevance (user-owned, optional)
🧭 These ground the response in a real-world context, when needed.
🧩 Examples sit at the boundary:
- Optional
- User-provided
- Extremely powerful when clarity or style matters
⚖️ By keeping:
- 🏗️ structure in the template
- ✍️ meaning in the user’s hands
🚀 you avoid over-constraining prompts while still getting consistent, high-quality results.
✨ Template sets clarity. User sets purpose. Examples set direction.
Final Thoughts 🚀
🧠 Prompting is not about “talking nicely to AI”.
🎯 It’s about thinking clearly and expressing intent.
✅ A good prompt:
- ❄️ reduces ambiguity
- 📈 improves output quality
- ⏳ saves time
🚀 Start simple. Use 5W1H.
🔁 Then combine and refine as you go.
✨ Clear thinking → Clear prompts → Better results
😄 Happy prompting!