Practical notes: Useful AI Agent Prompts Every Developer Should Have Saved

Stop typing the same prompts from scratch. Here’s a battle-tested library for your daily workflow.

Code Review & Quality

Deep code review

You are a senior software engineer doing a thorough code review.

Review the following code for:
- Correctness and logic errors
- Edge cases that aren't handled
- Security vulnerabilities (injection, auth issues, data exposure)
- Performance problems (N+1 queries, unnecessary loops, memory leaks)
- Readability and naming clarity
- Missing or misleading commentsFor each issue, state: the problem, why it matters, and a concrete fix.
Don't mention things that are fine. Only flag real issues.Code:
[PASTE CODE HERE]

Rubber duck a subtle bug

I have a bug I can't figure out. I'll describe what I expect to happen and what's actually happening. Ask me clarifying questions one at a time to help me find the root cause. Don't guess the answer yet — walk me through it.

Expected behavior: [DESCRIBE]
Actual behavior: [DESCRIBE]
What I've already tried: [DESCRIBE]Relevant code:
[PASTE CODE HERE]

Spot the performance bottleneck

Analyze this code for performance issues. Focus on:
- Time complexity of key operations
- Unnecessary re-computation or redundant work
- Memory allocation patterns
- Any blocking operations in an async context
- Database query efficiency (if applicable)

Suggest specific optimizations with expected impact. Show before/after where relevant.Language/framework: [e.g., Node.js / PostgreSQL]
Code:
[PASTE CODE HERE]

Writing & Improving Code

For the Writing Improving Code stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Implement a feature from a spec

For the Implement a feature from stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Implement the following feature in [LANGUAGE/FRAMEWORK].

Requirements:
- [REQUIREMENT 1]
- [REQUIREMENT 2]
- [REQUIREMENT 3]Constraints:
- Must integrate with: [EXISTING SYSTEM OR PATTERN]
- Must handle errors by: [e.g., throwing custom errors / returning Result types]
- Must be testable in isolation (no hidden dependencies)Do not add features beyond what's listed. Add inline comments only where the logic is non-obvious.

Refactor for readability without changing behavior

For the Refactor for readability without stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Prefer small, testable units over sprawling scripts. When a step fails, the failure should point at a single responsibility rather than a tangled pipeline. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Refactor the following code to improve readability. Rules:
- Do NOT change behavior or external API
- Rename variables and functions to be self-documenting
- Break up functions longer than 20 lines if it improves clarity
- Remove clever one-liners that sacrifice readability for brevity
- Add a brief comment above any function whose purpose isn't obvious from its name

Show the refactored version with a short explanation of the key changes made.Code:
[PASTE CODE HERE]

Convert callback-style code to async/await

For the Convert callback-style code to stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Treat this stage as a contract between inputs and validated outputs. Name the artifacts, define success checks, and refuse silent partial completion. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Convert the following callback-based Node.js code to use async/await with proper error handling. Preserve all logic exactly. Use try/catch blocks. If the original uses EventEmitters or streams that can't be straightforwardly promisified, flag them and suggest an approach.

Code:
[PASTE CODE HERE]

Write a utility function from description

For the Write a utility function stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Record timings and token or query cost next to functional results. Cost visibility early prevents surprise bills when the path moves from demo to shared environments. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Write a [LANGUAGE] utility function that:
- Does: [DESCRIBE WHAT IT SHOULD DO]
- Input: [DESCRIBE INPUT TYPE AND SHAPE]
- Output: [DESCRIBE OUTPUT TYPE AND SHAPE]
- Edge cases to handle: [LIST THEM, e.g., empty arrays, null values, negative numbers]

Include TypeScript types if applicable. Include 3–5 usage examples as comments below the function.

Testing

For the Testing stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Keep configuration outside application code. Environment files, secret stores, and feature flags belong in one place operators can audit without reading the whole graph. Prefer structured outputs with schema validation over free-form prose when the next step is code or a tool call.

Generate comprehensive unit tests

For the Generate comprehensive unit tests stage, define the inputs, the owner of the step, and the exit criteria before changing code. Operators should be able to re-run the step from a known checkpoint without guessing hidden state. Document the happy path and the recovery path together. Retries, human gates, and dead-letter handling are part of the product, not later polish.

Write unit tests for the following function using [JEST / VITEST / MOCHA / etc.].

Cover:
- The happy path
- All documented edge cases
- Invalid inputs (null, undefined, wrong types, out-of-range values)
- Any error conditions the function is supposed to throwUse descriptive test names that read as plain English. Group related tests with describe blocks. Mock external dependencies.Function to test:
[PASTE FUNCTION HERE]

Write integration tests for an API route

Write integration tests for this Express route using supertest and Jest.

Test:
- Successful response (correct status code and response shape)
- Validation errors (missing or invalid fields)
- Authentication/authorization failures
- Resource not found
- Any side effects (e.g., email sent, DB record created)Mock the database and any external services. Use beforeEach/afterEach for setup and teardown.Route handler:
[PASTE CODE HERE]

Generate test data / fixtures

Generate realistic test fixtures for the following data model. Create:
- 1 valid "happy path" example
- 1 example at boundary values (min/max lengths, edge dates, zero amounts)
- 1 example that should fail validation and why

Format as [JSON / TypeScript const / JS object]. Add a comment on each fixture explaining what scenario it represents.Data model / schema:
[PASTE SCHEMA HERE]

Debugging & Diagnosis

Explain a cryptic error message

I'm getting this error in my [LANGUAGE/FRAMEWORK] application. Explain:
1. What exactly this error means in plain English
2. The most common causes, ranked by likelihood
3. Step-by-step how to diagnose which cause applies to my situation
4. The fix for each cause

Error:
[PASTE FULL ERROR AND STACK TRACE HERE]Context (what I was doing when it happened):
[DESCRIBE]

Diagnose a memory leak

Help me diagnose a memory leak in my Node.js application.

Symptoms: [e.g., heap grows steadily under load, never GC'd, crashes after ~2 hours]
Environment: Node.js [VERSION], [FRAMEWORK]
What I've already checked: [LIST]Here are the relevant parts of the code:
[PASTE CODE]Walk me through:
1. What patterns in this code are likely culprits
2. How to confirm the leak with heap snapshots or --inspect
3. The fix

Understand unfamiliar code

Explain this code to me as if I'm a competent developer but unfamiliar with this codebase/library.

Tell me:
- What it does at a high level (1–2 sentences)
- What each major section does
- Any non-obvious patterns or idioms being used and why
- What I'd need to understand before safely modifying itCode:
[PASTE CODE HERE]

Architecture & Design

Design a system component

I need to design [COMPONENT NAME] for a [TYPE OF APPLICATION].

Context:
- Scale: [e.g., ~10k requests/day, 1M users]
- Existing stack: [e.g., Node.js, PostgreSQL, Redis, deployed on AWS]
- Key constraints: [e.g., must be horizontally scalable, strong consistency required]Give me:
1. A high-level design with the main components and how they interact
2. The data model (tables or document schemas)
3. The API surface (endpoints or function signatures)
4. Where the hard problems are and how you'd approach them
5. What you'd defer to a v2

Review an architecture decision

I'm deciding between two approaches for [PROBLEM].

Option A: [DESCRIBE]
Option B: [DESCRIBE]My context:
- Team size: [N developers]
- Expected scale: [DESCRIBE]
- Existing infrastructure: [DESCRIBE]
- Main concerns: [e.g., operational complexity, cost, latency]Compare them honestly. Don't hedge — give me a recommendation and explain the trade-offs I'd be accepting.

Design a database schema

Design a PostgreSQL schema for the following domain.

Domain description:
[DESCRIBE YOUR DATA AND RELATIONSHIPS]Requirements:
- [e.g., support multi-tenancy]
- [e.g., soft deletes]
- [e.g., audit log of all changes]
- [e.g., efficient queries for X and Y access patterns]Output:
- CREATE TABLE statements with constraints and indexes
- Brief explanation of key design decisions
- Any denormalization choices and why

Documentation

Write a README for a project

Write a README.md for the following project.

Project name: [NAME]
What it does: [DESCRIBE]
Tech stack: [LIST]
Target audience: [e.g., internal devs, open source contributors, end users]Include sections for: overview, prerequisites, installation, usage with examples, environment variables, running tests, and contributing. Use a practical, no-fluff tone. Don't add sections that would be empty.

Document a function or module

Write JSDoc / docstring documentation for the following code.

Include:
- A one-sentence summary of what it does
- @param tags with types and descriptions for every parameter
- @returns with type and description
- @throws for any errors it can raise
- A usage example in @exampleBe precise about types. If a parameter has constraints (e.g., must be positive, must be ISO 8601), document that.Code:
[PASTE CODE HERE]

Write a technical design doc (TDD)

Write a technical design document for the following feature.

Feature: [NAME AND BRIEF DESCRIPTION]
Author: [YOUR NAME]
Status: DraftSections to include:
1. Problem statement — what problem this solves and for whom
2. Goals and non-goals
3. Proposed solution — high-level design
4. Detailed design — data models, API changes, key logic
5. Alternatives considered and why they were rejected
6. Open questions
7. Security and privacy considerations
8. Rollout planContext to incorporate:
[PASTE ANY CONTEXT — TICKETS, DISCUSSIONS, EXISTING CODE]

Git & Workflow

Write a meaningful commit message

Write a commit message for the following change. Follow the Conventional Commits format (type(scope): description). Add a body paragraph explaining *why* the change was made, not just what changed. Keep the subject line under 72 characters.

Changes made:
[PASTE DIFF OR DESCRIBE THE CHANGE]

Write a pull request description

Write a pull request description for the following change.

Include:
- What this PR does (2–3 sentences)
- Why it's being done (the problem or requirement it addresses)
- How to test it manually
- Any risks or areas that need careful review
- Links to related issues or tickets: [LIST]Changes:
[PASTE DIFF SUMMARY OR DESCRIBE KEY CHANGES]

Summarize a PR for a non-technical stakeholder

Summarize the following code changes for a non-technical stakeholder. Avoid jargon. Focus on what changes from the user's perspective, what risk (if any) it introduces, and when it will be available.

Keep it under 5 sentences.Changes:
[PASTE PR DESCRIPTION OR DIFF SUMMARY]

Learning & Exploration

Explain a concept with a practical example

Explain [CONCEPT] to me. I'm a [junior/mid/senior] developer familiar with [RELEVANT BACKGROUND].

Use a concrete, practical example — not a toy one. Show me code I'd actually write in a real project. After the example, explain the underlying mechanism in plain language. Then tell me: when should I use this, and when should I avoid it?

Find what you don’t know about a topic

I know the basics of [TOPIC]. List 10 things about [TOPIC] that intermediate developers often don't know but should. For each one, give a one-paragraph explanation and a practical example of why it matters.

Prepare for a technical interview topic

I have an interview coming up and I want to study [TOPIC] deeply.

Give me:
1. The 10 most important concepts to understand
2. For each concept: a clear explanation, a code example, and a common interview question about it
3. The 3 questions most likely to trip up mid-level candidates on this topic, with ideal answers

Prompting Tips That Apply to All of These

Operational checklist