Engineering Prompts

This section contains 175+ prompts purpose-built for software engineering workflows. Every prompt has been tested against real codebases, across multiple LLMs, and refined until it consistently produces outputs that a senior engineer would accept without major revision.

Who This Is For

You write code for a living. You have opinions about error handling. You know what a flaky test feels like at 2 AM. These prompts are written by engineers who share that context — they assume you already know how to code and want the AI to operate at your level.

How These Prompts Are Organized

Each prompt entry follows a consistent structure:

### Prompt Title
**Category:** Where it fits in your workflow
**When to use:** The specific situation that calls for this prompt
**Prompt:**
> The actual text you send to the AI

**Example input:** What you would paste alongside the prompt
**Expected output quality:** What a good response looks like
**Variations:** Alternative phrasings for different contexts

The Five Subcategories

Debugging Prompts — 50+ prompts

The largest section, because debugging is where most engineering time goes. Covers:

• General debugging and root cause analysis
• Performance profiling and optimization
• Memory leak identification
• Race condition diagnosis
• Production incident triage
• Log and error trace analysis

Use these when you have a bug and need a second pair of eyes that has seen every Stack Overflow answer ever written.

Go to Debugging Prompts →

Architecture Review Prompts — 50+ prompts

For evaluating existing systems or reviewing proposed designs. Covers:

• System design evaluation
• Scalability analysis
• Security posture review
• API design assessment
• Database schema review
• Infrastructure and deployment review
• Code-level architecture review

Use these when you need a thorough review that goes beyond "looks good to me."

Go to Architecture Review Prompts →

Code Generation Prompts — 50+ prompts

For generating production-quality code, not toy examples. Covers:

• API endpoint generation (Express, Fastify, NestJS)
• Database migration and schema generation (Prisma, Drizzle, raw SQL)
• Validation schema generation (Zod, Yup, Joi)
• React component generation
• Test generation (unit, integration, e2e)
• Boilerplate and scaffolding
• TypeScript type generation
• Documentation generation

Use these when you know exactly what you want and need the AI to write it correctly on the first try.

Go to Code Generation Prompts →

Refactoring Prompts — 25+ prompts

For improving existing code without changing behavior. Covers:

• Method and class extraction
• Conditional simplification
• Duplication removal
• Naming improvements
• Complexity reduction
• Design pattern application

Use these when the code works but makes you wince every time you open the file.

Go to Refactoring Prompts →

Testing Prompts — 25+ prompts

For generating tests and improving test strategy. Covers:

• Unit test generation
• Integration test design
• Edge case identification
• Test data generation
• Test strategy and coverage review

Use these when you need to go from 40% test coverage to 90% without writing every test by hand.

Go to Testing Prompts →

Prompt Selection Guide

Not sure which prompt to use? Start here:

Situation	Start With
Something is broken and I do not know why	Debugging Prompts
I need to write new code quickly	Code Generation Prompts
The code works but is ugly or fragile	Refactoring Prompts
I need to review a PR or design doc	Architecture Review Prompts
I need more tests or better tests	Testing Prompts
A production alert just fired	Debugging > Production Incident Prompts
I am designing a new system	Architecture Review > System Design
I need to generate types from an API	Code Generation > TypeScript Types

Tips for Engineering Prompts

Always Include Your Tech Stack

The difference between a generic answer and a usable one is context. Always state:

• Language and version (TypeScript 5.3, Python 3.12, Go 1.22)
• Framework (Next.js 14, FastAPI, Gin)
• Database (PostgreSQL 16, MongoDB 7, Redis 7)
• Runtime (Node 20, Bun 1.1, Deno 2)

Paste Real Code, Not Descriptions

Instead of: "I have a function that fetches users and filters them" Do: Paste the actual function. The AI catches bugs in code, not in your summary of code.

Specify Output Constraints

• "Return only the modified function, not the entire file"
• "Include error handling for network failures and invalid input"
• "Use async/await, not callbacks or .then() chains"
• "Follow the existing code style — semicolons, single quotes, 2-space indent"

Chain Prompts for Complex Work

For a large feature, do not ask the AI to write everything at once. Instead:

1. Generate the database schema first
2. Generate the API route using the schema from step 1
3. Generate the validation layer referencing the route from step 2
4. Generate the tests using all of the above as context

Each step produces better output because it has focused context.

Quality Expectations

Every prompt in this section has been evaluated against these criteria:

• Correctness: The output compiles/runs without modification in 90%+ of cases
• Completeness: Edge cases, error handling, and types are included
• Style: Output follows conventional patterns for the given framework
• Teachability: The AI explains its reasoning when asked

If a prompt consistently underperforms against these criteria, it is either revised or removed.

Version History

Date	Change
2026-03-17	Initial release with 175+ prompts across 5 categories

---

"A great engineering prompt does not just describe what you want — it describes the constraints, the context, and the quality bar. The AI fills in the rest."