Module 02: Anatomy of a High-Performance Prompt | Hello Pharma AI Upskilling Program

🎯 Learning Objectives

Identify the 5 core components of any high-performance prompt
Write explicit instructions instead of relying on assumptions
Understand instruction hierarchy and what the model obeys first
Know when to use minimal vs rich prompts
Recognise and fix the most common failure patterns

1. The 5 Core Components

Every high-performance prompt contains some combination of five elements. Missing even one degrades the output.

Component	Answers	Example
Role	Who is responding?	"You are a regulatory auditor..."
Task	What needs to be done?	"Identify compliance gaps in this SOP"
Context	What background matters?	"The audience is fresh pharmacy graduates..."
Constraints	What are the limits?	"Max 150 words. No regulatory jargon."
Output Format	How should it look?	"Short intro + 5 bullet points"

2. Role — Who Is Responding

❌ No Role

"Explain tablets."

Generic output by necessity — no perspective assigned.

✅ With Role

You are a pharmaceutical manufacturing expert with experience in oral solid dosage forms. Explain tablets...

Useful Role Archetypes

Domain expert — "You are a senior regulatory affairs specialist..."
Adversarial — "You are an FDA inspector looking for compliance gaps..."
CXO advisor — "You communicate in plain executive language..."
Editorial — "You are a senior editor who prioritises clarity and precision..."

3. Task — What Exactly Needs to Be Done

❌ Vague Task

"Write about GMP."

Overview? Technical? Regulatory? Training material? The model picks one at random.

✅ Clear Task

Your task is to explain the importance of GMP to new operators joining a pharma manufacturing plant.

4. Context — Background & Audience

❌ Missing Context

"Explain validation."

For whom? A PhD scientist? A first-day operator? Each needs a completely different explanation.

✅ With Context

Context: The audience is fresh pharmacy graduates joining a manufacturing facility. They understand theory but lack practical exposure.

5. Constraints — Boundaries That Improve Quality

❌ No Constraints

"Explain cleaning validation."

You may get 800 words, regulatory citations, and tangential topics.

✅ With Constraints

Constraints: - Avoid regulatory citations - Focus on shop-floor relevance - Limit to 5 bullet points - No more than 150 words

6. Output Format — How the Answer Should Look

❌ No Format

"Explain process validation."

Wall of text — hard to scan, adapt, or share.

✅ With Format

Output Format: - Heading (1 line) - 5 bullet points (max 20 words each) - One practical example (2–3 sentences)

7. Weak vs High-Performance: Full Comparison

❌ Weak Prompt

"Explain GMP validation."

✅ High-Performance Prompt

You are a pharmaceutical quality consultant. Your task: Explain GMP validation to new manufacturing supervisors. Context: They have shop-floor experience but limited QA exposure. Constraints: Practical only · No regulatory jargon · Max 150 words Output: Short intro (2 sentences) + 5 bullet points

💡 Rule of Thumb

Higher stakes = richer prompt. A quick internal note can use a minimal prompt. A client deliverable or regulatory document deserves a fully structured prompt.

✏️ Module 02 Exercise

Take the weakest prompt you've used this week. Rebuild it using all five components: Role, Task, Context, Constraints, Output Format. Run both versions and compare the outputs.

🔑 Key Takeaways

Every prompt has 5 components — missing any one degrades output
Role primes perspective; Task drives direction; Context calibrates depth
Constraints reduce noise and improve usefulness
Format turns output into something immediately usable
Higher stakes demand richer, more structured prompts
Instruction conflicts produce diluted, generic outputs

← M01: Foundations of Prompt Engine… M03: Prompting Patterns & Framewo… →