Prompt Engineering: From Beginner to Expert

Bad prompt:  "Write about dogs"
Result:      Generic, unfocused content

Good prompt: "Write a 200-word guide for first-time dog owners 
             choosing between a Labrador and Golden Retriever, 
             comparing their exercise needs, temperament, and 
             grooming requirements. Use a friendly, helpful tone."
Result:      Exactly what you needed

[Role/Context] + [Task] + [Format] + [Constraints]

Example:
"You are a senior software engineer.        <- Role
 Review this code for security issues.      <- Task
 List each issue with severity and fix.     <- Format
 Focus on SQL injection and XSS only."      <- Constraints

❌ Vague: "Make it better"
✓ Specific: "Improve readability by adding comments, using descriptive 
             variable names, and breaking the function into smaller parts"

❌ Vague: "Write about AI"
✓ Specific: "Write a 500-word explanation of how neural networks learn,
             suitable for someone with basic programming knowledge"

❌ "Fix the bug in my code"
   (What code? What bug?)

✓ "This Python function should return the sum of even numbers, 
   but it's returning the sum of all numbers. Here's the code:
   
   def sum_evens(numbers):
       total = 0
       for n in numbers:
           total += n
       return total
   
   Fix the bug and explain what was wrong."

❌ "What's 23 × 47?"
   (AI might just guess)

✓ "What's 23 × 47? Think step by step."
   
   AI response:
   "Let me break this down:
   23 × 47 = 23 × (50 - 3)
           = 23 × 50 - 23 × 3
           = 1150 - 69
           = 1081"

"Convert these sentences to formal English:

Casual: 'gonna grab some food, wanna come?'
Formal: 'I am going to get some food. Would you like to join me?'

Casual: 'nah, im good'
Formal: 'No, thank you. I am fine.'

Casual: 'btw the meeting got pushed to 3'
Formal: "

"You are a patient, encouraging coding tutor who:
- Never gives complete solutions immediately
- Asks guiding questions to help students discover answers
- Celebrates small wins and progress
- Uses simple analogies to explain complex concepts

A student asks: 'I don't understand recursion'"

"Analyze this product review and respond in JSON:
{
  'sentiment': 'positive/negative/neutral',
  'confidence': 0.0 to 1.0,
  'key_points': ['point1', 'point2'],
  'improvement_suggestions': ['suggestion1']
}

Review: 'Love this product but shipping was slow'"

response = openai.chat.completions.create(
    model="gpt-4",
    messages=[
        {"role": "system", "content": "You are a helpful assistant..."},
        {"role": "user", "content": "What's the weather?"}
    ]
)

## Core Identity
You are [name], a [role] that helps users [primary goal].

## Personality
- [Trait 1]: description
- [Trait 2]: description

## Key Behaviors
1. Always [behavior]
2. Never [anti-behavior]
3. When [situation], do [action]

## Response Format
- Keep responses [length guideline]
- Use [formatting preference]
- Include [required elements]

## Boundaries
- Don't [restriction 1]
- Don't [restriction 2]

## Core Identity
You are Aria, a friendly customer support agent for TechCorp, 
a software company that makes project management tools.

## Personality
- Warm and empathetic: Acknowledge frustrations before solving
- Clear communicator: No jargon, step-by-step instructions
- Proactive: Anticipate follow-up questions

## Key Behaviors
1. Always greet users warmly
2. Always confirm you understood their issue
3. Never blame the user for problems
4. When you can't help, escalate to human support
5. End interactions by asking if there's anything else

## Knowledge
- You can help with: account issues, billing, feature questions
- You cannot: process refunds (must escalate), access user data

## Response Format
- Keep responses under 150 words unless explaining complex steps
- Use numbered lists for multi-step instructions
- Bold key actions: "Click **Settings** then **Account**"

C - Capacity: What role should the AI assume?
R - Request: What specific task should it do?
I - Information: What context does it need?
S - Style: How should it communicate?
P - Preferences: Any special requirements?
E - Examples: Sample inputs/outputs if helpful

Example:
"Capacity: You are a data scientist
Request: Analyze this dataset for trends
Information: Sales data from 2020-2023, columns: date, product, revenue
Style: Professional but accessible
Preferences: Focus on quarterly patterns
Examples: 'Q1 2022 showed 15% growth compared to Q1 2021'"

Prompt 1: "Write about machine learning"
(Too broad, not useful)

Prompt 2: "Write an introduction to machine learning for business executives"
(Better, but still vague)

Prompt 3: "Write a 300-word introduction to machine learning for 
          business executives who need to decide whether to invest 
          in ML projects. Focus on practical applications and ROI."
(Specific and actionable)

"Important constraints:
- Response must be under 100 words
- Use only vocabulary a 10-year-old would understand
- No technical jargon

Now explain: What is a computer virus?"

"Write a Python function that:
- Name: validate_email
- Input: string (email address)
- Output: boolean (True if valid)
- Requirements:
  * Check for @ symbol
  * Check for domain with dot
  * Handle edge cases (empty string, None)
- Style: Include docstring, type hints
- Don't use: regex (use string methods only)"

"Write a blog post:
- Topic: Benefits of remote work
- Audience: HR managers considering hybrid policies
- Tone: Professional but engaging
- Length: 800 words
- Structure: 
  * Hook opening with a statistic
  * 3 main benefits with examples
  * Address one common concern
  * Call to action
- SEO: Include keywords 'remote work benefits', 'hybrid workplace'"

"Analyze this sales data and provide:
1. Key trends (3-5 bullet points)
2. Anomalies that need investigation
3. Recommendations (prioritized by impact)

Data context:
- Monthly sales figures for 2023-2025
- Products: A, B, C
- Regions: North, South, East, West

Focus on: Seasonal patterns and regional differences
Format: Markdown with headers"

❌ "Help me with my project"
✓ "I'm building a React app for a todo list. I need help 
   implementing drag-and-drop reordering of tasks."

❌ "List the pros and cons of Python"
✓ "Create a comparison table of Python pros and cons with 
   columns: Category, Advantage/Disadvantage, Example"

❌ "Be concise but also be thorough and explain everything in detail"
✓ "Provide a concise summary (2-3 sentences), then a detailed 
   explanation (2-3 paragraphs) for those who want more depth"

❌ "Continue the code from before"
   (New session, AI doesn't remember)
✓ "Here's my existing code: [paste code]. Add a function 
   that validates user input."