M4-A — Reusable Prompt Systems
In Module 3 you learned how to write a good prompt once. In Module 4 you learn how to save that prompt and use it again and again — like saving a recipe instead of cooking from memory every time.
🍳 The problem with "one-time" prompts
Imagine you ask ChatGPT to write a WhatsApp reply to a student enquiry.
Today it writes something great.
Tomorrow you ask the same thing — different tone, different length, different format.
That is not your fault. That is what happens when every chat starts from zero.
| One-time prompt | Reusable prompt system |
|---|---|
| You rewrite instructions every time | Instructions are saved once |
| Output quality jumps up and down | Output stays more consistent |
| Only you remember what worked | Your whole team can use the same file |
| Lives inside chat history | Lives in a .md file you control |
A prompt template is not magic words. It is a saved instruction sheet that tells AI exactly how to behave for one type of job.
📁 What is a reusable prompt system?
A reusable prompt system has three parts:
1. Template file → The instructions (usually a .md file)
2. Your input → The new data each time (name, product, question)
3. AI output → The answer in the format you asked for
Think of it like a form at a coaching class:
- The form already has fixed fields: Name, Phone, Course, Batch
- You only fill in today's student details
- The office gets the same structured information every time
Your prompt template works the same way.
🧩 Parts of a good prompt template
Copy this structure into any .md file in Obsidian:
| Section | What it does | Example |
|---|---|---|
| Role | Who should AI act as? | "Act as a friendly coaching counsellor" |
| Goal | What job must it finish? | "Write a WhatsApp reply to a new enquiry" |
| Context | Background rules | "We teach AI courses in Mumbai. Tone is warm but professional." |
| Input slots | Blanks you fill each time | Student name: ___ / Question: ___ |
| Output format | How answer should look | "Max 4 lines. No emojis. Hindi-English mix okay." |
| Do NOT rules | What to avoid | "Do not promise discounts. Do not invent course dates." |
Make a folder like My Prompt Templates/ and keep one file per job: whatsapp-reply.md, linkedin-post.md, exam-summary.md.
🔄 How to use a template (step by step)
Step 1: Open your template file
Step 2: Copy the whole template
Step 3: Paste into ChatGPT / Claude / Gemini
Step 4: Fill in today's input slots
Step 5: Send → get answer in same format every time
Example — Student doubt reply template
Role: You are a patient AI tutor for Class 12 students.
Goal: Answer one student doubt in simple language.
Context: Student just passed school. No coding background.
Input:
- Subject: {{subject}}
- Doubt: {{question}}
Output format:
- 1 short paragraph answer
- 1 real-life analogy
- 1 "remember this" line at the end
- No emojis
Do NOT:
- Use heavy technical words without explaining them
- Give a 10-page essay
Each time you only change {{subject}} and {{question}}.
🏗️ From one prompt → a small AI system
When you have many templates for one workflow, you have a prompt system:
Enquiry comes in
↓
Template A → classify the question type
↓
Template B → draft the reply
↓
Template C → shorten for WhatsApp
↓
Human reads and sends
You are not "chatting with AI" anymore.
You are running a small process made of saved prompts.
That is what the brochure means by "Reusable AI systems".
🎯 Where students should use this
| Situation | Template name idea |
|---|---|
| Daily study notes | chapter-to-notes.md |
| Instagram caption | product-caption.md |
| Email to teacher | formal-email.md |
| Resume bullet points | resume-achievement.md |
| Project report intro | report-intro.md |
When a template stops giving good answers, edit the file. Better templates come from real use — not from writing once and never opening again.
✅ Checklist — did you build a real system?
Next: M4-B — Prompt-Based Automation — connect templates into automatic workflows.
Also see: Ruleset Prompt — rules that control how AI behaves in every chat.