Production-grade PRDs that let AI agents one-shot entire applications.

What Is This?

This is a collection of extremely detailed, battle-tested Product Requirements Documents - each one 10,000+ lines - designed to be fed directly to an AI agent so it can build a complete, production-ready application in a single shot.

No back-and-forth. No "can you also add...". No debugging prompt misunderstandings.

One PRD. One agent session. One fully working application.

This repo also includes the skills and methodology we use to produce PRDs at this bar - so you can write your own.

Available PRDs

PRD	Description	Lines	Tech Stack
Lightweight CRM	Full-featured CRM with contacts, companies, deals pipeline, activities, email integration, dashboards, and role-based access control	10,000+	React · Node.js · PostgreSQL · Tailwind CSS

More PRDs coming soon - task management, company knowledge base, and more.

Available Skills

Skills are reusable methodologies that AI agents can follow to produce work at a consistent bar. They're how we make PRDs like the ones above.

Skill	Description
PRD Builder	The complete methodology for producing AI-agent-executable PRDs - seven phases, dual review, three drafts, environment-agnostic

Hand a skill to your agent (paste its contents, or load it as a skill in your platform) and the agent will follow the methodology to produce a PRD at the same bar as the ones in this repo.

How To Use

With Any AI Agent

Copy the contents of a PRD file
Feed it to your AI coding agent (Claude, GPT, Cursor, etc.)
The agent builds the entire application

The PRDs are written to be self-contained and unambiguous - every database schema, every API endpoint, every UI component, every edge case is specified. A capable AI agent can execute them without asking follow-up questions.

With CRHQ

These PRDs were built on and optimized for CRHQ - an AI agent deployment platform where autonomous agents run on dedicated servers with real tools.

On CRHQ, the workflow is:

Feed the PRD to your Project Manager agent
The PM coordinates multiple Developer agents to build it and deploy it, while Tester agents run end-to-end tests using scripts and a headless browser
You get a live URL with a working, tested application

No local setup. No terminal juggling. No deployment pipeline. The agents have a full Linux server, database, file system, and browser - they handle everything.

Why it works better on CRHQ:

Persistent memory - the agent remembers every decision across sessions
Real tools - file system, PostgreSQL, headless browser, cron jobs, credential vault
Skills system - coding standards, UI guidelines, and best practices are enforced automatically
Auto-deployment - what gets built goes live immediately
No context window limits - 10,000+ line PRDs execute without truncation

Get started with CRHQ →

What Makes These PRDs Different

Most PRDs are vague wishlists. These are executable specifications.

Every database table is fully defined - columns, types, constraints, indexes, seed data
Every API endpoint is specified - method, path, request/response schemas, error codes, auth rules
Every UI component is described - layout, interactions, responsive behavior, empty states, loading states
Every edge case is handled - validation rules, error messages, race conditions, permission boundaries
Every configuration is included - environment variables, third-party integrations, deployment steps

An AI agent reading one of these PRDs has zero ambiguity about what to build.

Who Is This For

Solo founders who need production applications without a dev team
CTOs evaluating AI-driven development workflows
Developers who want a reference for how to write PRDs that AI can execute
AI enthusiasts pushing the boundaries of what agents can build autonomously

About CRHQ

CRHQ is an AI agent deployment platform. Every user gets a dedicated cloud server - a Satellite - where autonomous AI agents run 24/7 with real tools: file systems, databases, browsers, scheduled jobs, persistent memory, and an encrypted credential vault.

We don't build chatbots. We build employees.

One person orchestrating a team of CRHQ agents outperforms a department 10x its size. These PRDs are proof - a single agent, given the right instructions, can build what would take a small team weeks.

Deploy your first agent in 90 seconds →

Contributing

We welcome contributions of both PRDs and skills.

For PRDs (prds/):

5,000+ lines minimum - we're not interested in toy specs
Fully executable - an AI agent should be able to build the entire application from your PRD alone
Battle-tested - the PRD must have been used to successfully generate a working application at least once
Self-contained - no external docs, no "see also", no assumptions

For skills (skills/):

Tech-stack agnostic - the methodology should work across stacks and project types
Battle-tested - the skill must have been used by an agent to produce a real artifact at least once
Self-contained - no environment-specific assumptions; reference the host's facilities generically
Folder per skill - skills/<skill-name>/SKILL.md plus any supporting scripts, templates, or examples

Open a PR and we'll review it.

License

MIT - use these PRDs however you want. Build with them, learn from them, fork them.

Built by CRHQ - Autonomous AI agents on dedicated servers.

CRHQ-PRDs