Prompt Contracts book cover
The Book

Prompt Contracts: The Book

How I Stopped Vibe Coding and Started Shipping Real Software With AI

A field manual from a solo developer who ships production software with AI every single day. Not demos. Not proofs of concept. Real products, with real users, paying real money.

Free with Kindle Unlimited

Or buy it: Kindle $9.99

A prompt contract is a one-page specification you write before handing a task to an AI coding agent like Claude Code, Cursor, or GitHub Copilot. Instead of hoping the AI understands your intent, you define the stack, constraints, inputs, outputs, and acceptance criteria upfront. The result: production-ready code on the first pass, not endless debugging cycles. This book teaches the complete framework in 9 chapters with 10 copy-paste templates and runnable examples you can use today.

The Medium article that started it all

46K
Views
29K
Reads
+418
Subscribers
+538
Followers

63% read ratio · 16.8K from Google · 62% programmers, 48% AI practitioners

What readers say

"The better your specs are, the better your product is."

Randall Bownds

"It will build what you say. Do you say what you need?"

Vitaly Hornik

"It's good to know that the practices you encountered are exactly the same ones I use."

artful_dodger 71 claps

"Planning before execution is key."

Manu Gupta 159 claps

"This is an important article. Writing requirements was the precursor to design and coding phases."

Joe Bologna

"Basically it's not possible to vibe code without already knowing how to code?"

Money, Geopolitics & Oil 117 claps

From the 67 responses on Medium

What's inside the prompt contracts book

The complete prompt contract framework with copy-paste templates
A full tech stack walkthrough with honest assessments
The verification loop: AI tests its own code before you take delivery
A weekend SaaS build from zero to first paying customer (95% margins)
Replace a $200/month AI subscription with a $5/month self-hosted setup
Five automation patterns that survive production
Why Lovable and Bolt produce toys, not products
The honest truth about shipping solo: burnout, scope creep, money

9 chapters, 260 pages, zero filler

0
The Gap Between Demo Code and Production Code
Stop wasting days debugging AI code that looked right but breaks in prod
1
Prompt Contracts: The Framework
Write a one-page spec, get production-ready code on the first pass
2
The Stack and the Brain
Pick a stack that works together and configure CLAUDE.md so your AI knows your codebase
3
The Verification Loop
Make your AI test its own code before you even look at it
4
Ship a SaaS in a Weekend
Follow a complete build from zero to first paying customer with 95% margins
5
Self-Host Your AI Agent for $5/Month
Cut your AI bill from $200/month to $5 and own your infrastructure
6
Automate Like a Maniac
Build workflows that run your business while you sleep
7
Lovable and Bolt Are Toys
Know when to prototype fast and when to build for real
8
Shipping Solo: The Real Game
Survive burnout, scope creep, and loneliness — from someone who does it daily
"The specification is the product. The code is the delivery."

About the author

Phil writes on Medium under @rentierdigital. 20 years in software, from French banks to a shrimp fishery in Madagascar. Now a solo developer shipping SaaS products with AI from wherever the Wi-Fi holds.

Stack: Next.js, Convex, Clerk, Vercel, n8n. Self-hosted AI agents. Everything in this book has been deployed, broken in production, fixed, and deployed again.

Every chapter comes with runnable code

Scan the QR codes inside the book to access templates, workflows, and companion repos

Companion Repository Ch.0 + Ch.8

Templates, runnable examples, and the full tool stack from the book.

10 templates · 6 runnable examples · full stack breakdown

Prompt Contract Templates Ch.1 + Ch.3

10 ready-to-use templates: API routes, database schemas, React components, auth flows, debugging.

Copy a template → fill the blanks → hand to Claude Code → get working code

The Full Stack Ch.2 + Ch.5

Every tool mentioned in the book with setup links, cost breakdowns, and honest comparisons.

Convex · Clerk · Next.js · n8n · Hostinger · and more

Real CLAUDE.md Examples Ch.2

Production CLAUDE.md files from real projects. See how to configure Claude Code for a SaaS app and an automation pipeline.

SaaS project config · Automation project config

DreamMiner — Complete SaaS Example Ch.4

The full contracts and code from Chapter 4. Stripe webhooks, Convex mutations, auth pages — all built contract-first.

4 contracts · Convex schema · Stripe webhook · Next.js pages

QR CLI Tool Ch.6

A batch QR code generator built from a prompt contract. Python CLI with sample CSV and test script. Run it in 2 minutes.

Python · pip install → run → QR codes generated

n8n Workflow Contracts Ch.6

Prompt contracts for n8n workflows: content pipelines, cross-platform syndication, monitoring alerts.

3 contracts · feed to Claude Code + n8n-mcp → workflows built on your instance

Prototyping Tools Ch.7

The tools I use to validate ideas in 20 minutes before building properly. Lovable, v0, and the honest assessment.

Validate fast · build properly · no-code for prototypes, contracts for production

Your AI has been waiting for better instructions.

Get Prompt Contracts on Amazon

Free with Kindle Unlimited

Or buy it: Kindle $9.99

Frequently asked questions

What is a prompt contract?

A prompt contract is a structured specification you write before giving a task to an AI coding agent. It defines the tech stack, inputs, outputs, constraints, and acceptance criteria so the AI delivers working code on the first pass instead of generating throwaway demos.

Do I need coding experience to use this book?

You need enough technical understanding to describe what you want built. You don't need to write code yourself, but you need to understand concepts like APIs, databases, and deployment. The book covers how to specify requirements clearly enough that AI does the implementation.

How is this different from vibe coding?

Vibe coding is typing a loose prompt and hoping the AI figures it out. It works for demos but breaks in production. Prompt contracts flip the process: you invest 10 minutes writing a clear spec, and the AI delivers production-grade code that handles edge cases, error states, and real-world constraints.

Which AI tools does this work with?

The examples use Claude Code, but the prompt contract framework works with any AI coding agent: Cursor, GitHub Copilot, Windsurf, or ChatGPT. The principle is the same: better input specifications produce better output code, regardless of the tool.