Prompt Engineering

Core Techniques

Be Specific and Structured

BAD: "Fix this code"
GOOD: "Review this Python function for:
1. Security vulnerabilities
2. Performance issues
3. Error handling gaps
Return findings as a numbered list with file:line references."

Provide Context

BAD: "Write a REST API"
GOOD: "Create a Express.js REST API for a todo app with:
- TypeScript
- PostgreSQL via Prisma
- JWT authentication
- CRUD endpoints for todos
- Input validation with Zod"

Chain-of-Thought

Think step by step:
1. Analyze the requirements
2. Identify edge cases
3. Design the solution
4. Implement the code
5. Verify correctness

Prompt Templates

Code Review

Review the following code for:
- Bugs and logic errors
- Security vulnerabilities
- Performance bottlenecks
- Code style and readability
- Missing error handling

Provide specific line references and concrete fix suggestions.

Code:

Refactoring

Refactor this code to:
- Improve readability
- Reduce duplication
- Follow SOLID principles
- Add type safety

Maintain the same public API and behavior.
Show the refactored code with a brief explanation of changes.

Code:

Test Generation

Generate comprehensive tests for this function:
- Cover happy path
- Cover edge cases (empty input, null, boundary values)
- Cover error cases
- Use the same testing framework as the project
- Include descriptive test names

Function:

Debugging

I'm getting this error: [ERROR MESSAGE]

Context:
- Language/framework: [LANGUAGE]
- What I expected: [EXPECTED]
- What actually happens: [ACTUAL]
- Relevant code: [CODE]

Help me identify the root cause and fix it.

Documentation

Generate documentation for this code:
- Brief summary of what it does
- Parameters/arguments with types
- Return value
- Usage example
- Any important notes or caveats

Code:

Few-Shot Examples

Convert these function calls to SQL:

Input: getUser(123)
Output: SELECT * FROM users WHERE id = 123;

Input: getPostsByAuthor(authorId=456, limit=10)
Output: SELECT * FROM posts WHERE author_id = 456 LIMIT 10;

Input: updateUserEmail(userId=789, email="new@example.com")
Output: UPDATE users SET email = 'new@example.com' WHERE id = 789;

Now convert:
Input: searchProducts(category="electronics", minPrice=100, maxPrice=500)
Output:

System Prompt Patterns

Role Assignment

You are a senior software engineer with expertise in Python and system design.
You write clean, well-documented code following PEP 8.
You consider edge cases and error handling in every solution.

Output Format Control

Always respond in this format:
## Analysis
[Brief analysis of the problem]

## Solution
[Code implementation]

## Explanation
[1-3 sentences explaining the approach]

## Alternatives
[Other valid approaches, if any]

Constraint Setting

Constraints:
- Use only standard library (no external dependencies)
- Target Python 3.10+
- Keep functions under 20 lines
- Include type hints
- No comments explaining obvious code

Advanced Patterns

Self-Consistency

Ask the same question multiple times, take the majority answer.

Tree of Thought

Consider three different approaches to solve this:
Approach 1: [describe]
Approach 2: [describe]
Approach 3: [describe]

For each, evaluate:
- Pros and cons
- Complexity
- Maintainability

Then recommend the best approach.

Reflection

Review your previous answer. Check for:
1. Correctness - are there any bugs?
2. Completeness - did you miss any requirements?
3. Efficiency - can it be optimized?
Provide a corrected version if needed.

Anti-Patterns

Don't	Do
"Make it good"	Specify what "good" means
Dump code without context	Explain the goal and constraints
One giant prompt	Break into steps
Assume the model knows your project	Provide relevant code/context
Ignore output format	Specify exact format needed

Token Optimization

• Put the most important instructions first
• Use concise language without losing meaning
• Reference code by file path, don't paste entire files
• For large codebases, specify which files to focus on
• Use structured formats (lists, tables) for clarity