AI Prompt Engineering for Developers

Effective prompting is the difference between getting useful code from an AI assistant and getting garbage. This guide covers practical techniques for developer workflows.

Core Principles

Be Specific

Bad: "Write a function to handle users"

Good: "Write a TypeScript function that validates an email address using a regex, returns true for valid emails and false otherwise, and handles edge cases like plus-addressing and subdomains."

Provide Context

Tell the model what you're working with:

Language and framework

Existing code patterns

Constraints (performance, browser compatibility)

What you've already tried

Show Examples

Few-shot prompting dramatically improves output quality:

Convert these SQL queries to Prisma:

SQL: SELECT * FROM users WHERE age > 21
Prisma: await prisma.user.findMany({ where: { age: { gt: 21 } } })

SQL: SELECT * FROM posts WHERE authorId = 5 ORDER BY createdAt DESC LIMIT 10
Prisma:

Techniques

Chain of Thought

Ask the model to think step by step for complex problems:

"Before writing the code, outline the algorithm steps. Then implement each step with comments."

Role Assignment

"You are a senior TypeScript developer who prioritizes type safety and writes clean, well-tested code."

Iterative Refinement

Start broad, then refine:

"Generate a REST API for a todo app"

"Add input validation using Zod"

"Add rate limiting middleware"

"Add OpenAPI documentation"

Negative Constraints

Tell the model what NOT to do:

"Don't use any external dependencies. Don't use var. Don't include console.log statements."

Common Patterns for Coding

Code Review

"Review this code for security vulnerabilities, performance issues, and adherence to TypeScript best practices. For each issue, explain the risk and provide a fix."

Test Generation

"Write unit tests for this function using Vitest. Cover edge cases including empty input, invalid types, and boundary values. Use describe/it blocks."

Documentation

"Write JSDoc comments for this module. Include @param, @returns, @throws, and @example tags."

Building Prompts Programmatically

Use our Prompt Builder tool to construct multi-message prompts with system/user/assistant roles, template variables, and export to JSON format compatible with OpenAI and Anthropic APIs.

Conclusion

Prompt engineering is a practical skill. Be specific, provide context, use examples, and iterate. The better your prompts, the more useful AI becomes in your daily workflow.