Prompting Guide for Developers

Prompts for Writing Code

AI doesn''t replace the developer — it makes them faster and more efficient. A good code generation prompt can save you hours of development time, provided you know how to communicate precisely what you want.

Principles of an Effective Code Prompt

The golden rule: the more specific your prompt, the better the generated code. Always include:

• Language and framework: "in TypeScript with Next.js 14" — not just "in JavaScript"
• Project conventions: "using named exports, interfaces over types, and @/* import aliases"
• Technical context: dependency versions, existing architecture, patterns used
• Edge cases: error handling, null/undefined cases, input validation
• Performance constraints: "must handle 10K requests/second", "optimized for bundle size"

React Component Template Prompt

Create a React component in TypeScript for [functional description]. Stack: Next.js 14, Tailwind CSS 4, next-intl for i18n. The component should: 1) use Server Components by default (add ''use client'' only if necessary), 2) type all props with an exported interface, 3) handle loading and error states, 4) be accessible (ARIA labels, keyboard navigation), 5) be responsive (mobile-first). No default exports.

API Endpoint Template Prompt

Create a REST API endpoint in [language/framework] for [functionality]. The endpoint must: 1) validate inputs with [Zod/Joi/other], 2) handle authentication via [method], 3) return appropriate HTTP codes (200, 400, 401, 404, 500), 4) include exhaustive error handling with explicit messages, 5) log errors with debugging context, 6) be documented with JSDoc comments.

---

Debugging with AI

Debugging is perhaps the use case where AI shines most for developers. It can analyze stack traces, identify subtle bugs, and propose fixes in seconds.

Error Analysis Prompt

I have this error in my [language/framework] application:
[paste complete error with stack trace]
Context: [describe what the code should do, when the error occurs].
Relevant code: [paste code].
1) Explain the likely cause, 2) Propose a fix with modified code, 3) Explain why the fix resolves the issue, 4) Identify any similar problems in the shown code.

Intermittent Bug Prompt

I have an intermittent bug in [context]. It manifests as [symptoms] and occurs approximately [frequency]. Environment: [prod/staging/dev]. Here''s what I''ve already checked: [list]. Here are the relevant logs: [logs]. Propose an ordered list of hypotheses (most to least probable) with for each: 1) why this hypothesis, 2) how to verify it, 3) how to fix if confirmed.

---

Refactoring and Code Review

AI is an excellent refactoring partner. It can analyze legacy code, identify code smells, and propose improvements while preserving functional behavior.

Complete Refactoring Prompt

Here is a module/component/function I want to refactor:
[paste code]
Constraints: 1) functional behavior must remain identical, 2) improve readability and maintainability, 3) follow SOLID principles, 4) extract magic constants, 5) improve variable and function naming, 6) add missing TypeScript types. Provide: the complete refactored code, and a summary of changes with justifications.

Code Review Prompt

You are a senior developer doing a thorough but constructive code review. Review this code:
[paste code or diff]
Evaluate on: 1) Correctness, 2) Security (injection, XSS, etc.), 3) Performance, 4) Maintainability, 5) Test coverage. For each issue, indicate severity (critical/major/minor/suggestion) and propose a fix.

---

Automated Testing

Writing tests is often the least enjoyed part of development, but it''s also where AI can deliver considerable value.

Unit Testing Prompt

Write comprehensive unit tests for this function/module:
[paste code]
Test stack: [Vitest/Jest/Pytest/other]. Cover: 1) all nominal cases (happy path), 2) edge cases (empty inputs, null, undefined, very large numbers, special strings), 3) error cases (invalid inputs, network errors, timeouts), 4) concurrency cases if applicable. Use descriptive test names. Organize tests with describe/it logically. Mock external dependencies.

E2E Testing Prompt

Write an E2E test with [Playwright/Cypress] for the following user journey: [step-by-step description]. The test should: 1) use robust selectors (data-testid preferred), 2) explicitly wait for elements (no sleep), 3) verify each step, 4) take screenshots at key steps, 5) handle failure cases gracefully.

---

Architecture and Documentation

AI can help you make informed architectural decisions and produce quality technical documentation — two activities often neglected due to time constraints.

Architecture Design Prompt

I need to design the architecture for [feature/system]. Constraints: 1) Tech stack: [stack], 2) Expected load: [metrics], 3) Infra budget: [budget], 4) Team: [size and seniority], 5) Timeline: [timeline]. Propose 2-3 architectural approaches with pros, cons, implementation complexity, estimated cost, and scalability for each. Recommend your preferred choice with justification.

---

Developer + AI Best Practices

What AI Does Well

• Boilerplate and repetitive code: CRUD, migrations, configurations, types
• Language/framework translation: Python to TypeScript, REST to GraphQL
• Code comprehension and explanation: legacy code, complex algorithms
• Test generation: from existing code, it identifies cases to test
• Documentation: JSDoc, README, ADR, code comments

What AI Does Poorly (or Dangerously)

• Large-scale system architecture: it lacks holistic vision of your codebase
• Security-critical code: authentication, cryptography — always have an expert review
• Advanced performance optimization: subtle optimizations require real profiling
• Business decisions: which feature to build, which tech debt to pay — that''s your job

Security Rules to Follow

• Never paste secrets (API keys, tokens, passwords) into a prompt
• Always review generated code before merging
• Don''t trust suggested dependencies: verify they exist, are maintained, and have no known CVEs
• Watch for API hallucinations: AI can invent methods or parameters that don''t exist

Go further with our developer prompt library and practice with our technical prompting exercises. Also discover how AI agents can automate your development workflows.