Table of Contents
- What is DevinOS?
- Why DevinOS?
- Quick Start
- Core Components
- Engineering Constitution
- Priority Hierarchy
- 67 Skills
- 22 Immutable Rules
- Self-Improving Knowledge Engine
- Repository Structure
- Quality Gates
- Contributing
- Roadmap
- License
What is DevinOS?
DevinOS is the world's first open-source Engineering Operating System built specifically for Devin and AI coding agents. It is a comprehensive, modular, and self-improving knowledge base that turns every AI-assisted project into a learning opportunity.
Think of it as the "brain upgrade" for Devin: a structured collection of rules, skills, workflows, playbooks, templates, memory, and prompts that guide AI agents to write production-grade software with the discipline of a senior engineer.
Unlike generic prompt libraries, DevinOS is:
- Constitution-Driven - Every rule derives from the Engineering Constitution
- Self-Improving - Learns from every project and grows smarter over time
- Production-Ready - Built for real-world software engineering, not toy examples
- Community-Powered - Federated learning from thousands of Devin users worldwide
Why DevinOS?
| Without DevinOS | With DevinOS |
|---|---|
| Generic, inconsistent output | Constitution-driven, principled decisions |
| Reinvents the wheel every project | Reuses battle-tested patterns and skills |
| Security overlooked | Security-by-default in every line |
| No memory between sessions | Growing knowledge base from every project |
| No quality standards | Rigorous quality gates for every contribution |
| AI hallucinations unchecked | Verification-before-declaration mandatory |
Quick Start
1. Install DevinOS in Any Project (30 seconds)
# Clone DevinOS
git clone https://github.com/Devin-IQ/DevinOS.git
# Enter your project folder
cd your-project
# Copy the brain into your project
cp -r ../DevinOS/.agents ./
# Done. Devin will auto-discover and use the skills and rules.
2. Global Install (IDE Integration)
Installs directly into your IDE's global Rules & Skills folders — no project setup needed.
Windows (Windsurf / Cursor):
irm https://raw.githubusercontent.com/Devin-IQ/DevinOS/main/install.ps1 | iex
Mac / Linux (Windsurf / Cursor / Claude Code):
curl -sL https://raw.githubusercontent.com/Devin-IQ/DevinOS/main/install.sh | bash
The installer auto-detects your IDE and copies:
- 21 Rules → Global Rules folder
- 40 Skills → Global Skills folder
Restart your IDE to activate.
Core Components
DevinOS Brain Structure
|
|-- ENGINEERING_CONSTITUTION.md # The Supreme Law
|-- .agents/
| |-- skills/ # 67 actionable capabilities
| |-- rules/ # 22 immutable principles
| |-- workflows/ # End-to-end processes
| |-- playbooks/ # Step-by-step guides
| |-- memory/ # Growing knowledge base
| |-- templates/ # Implementation starters
| |-- prompts/ # Curated prompt library
|
|-- docs/ # Deep documentation
|-- examples/ # Working code samples
|-- assets/ # Media and diagrams
Engineering Constitution
The Engineering Constitution is the supreme law of DevinOS. Every skill, rule, workflow, and memory entry must derive from it.
The 10 Articles
- Priority Hierarchy - Correctness > Security > Maintainability > Performance > DX > Readability > Speed
- Never Assume - Verify requirements, code, libraries, tests, documentation
- Verification Before Declaration - No task is complete without evidence
- Reuse Over Reinvention - Search existing code before creating new files
- Simplicity - Keep functions small, files focused, prefer composition
- Security by Default - Validate inputs, escape outputs, never expose secrets
- Continuous Learning - Extract generalizable lessons from every project
- Quality Gates - Uniqueness, documentation, examples, anti-patterns, verification
- Git Ethics - Small focused commits, meaningful messages, no secrets
- Human-AI Collaboration - AI augments humans; decisions must be explainable
Priority Hierarchy
All DevinOS content follows this immutable priority order:
1. Correctness -- It must work
2. Security -- Protect data and users
3. Maintainability -- Easy to understand and modify
4. Performance -- Fast enough for requirements
5. Developer Experience -- Reduce friction
6. Readability -- Clear and expressive
7. Speed -- Deliver quickly
No feature, optimization, or shortcut may violate this hierarchy.
67 Skills
Every skill includes frontmatter with name and description, then Description, Purpose, Trigger, Context, Workflow, Examples, Anti-patterns, Verification Checklist, and References.
Core Engineering
| Skill | Category | Purpose |
|---|---|---|
| Planning | Core | Break complex tasks into small, verifiable steps |
| Architecture Review | Core | Evaluate and improve architectural decisions |
| Structured Debugging | Core | Systematic, evidence-based bug fixing |
| Code Review | Core | Evaluate changes for correctness and security |
| Refactoring | Core | Improve structure without changing behavior |
| Testing | Core | Write comprehensive, deterministic tests |
| Verification | Core | Rigorous validation before declaring completion |
Specialized Disciplines
| Skill | Category | Purpose |
|---|---|---|
| Security Review | Security | Systematic vulnerability evaluation |
| Performance Optimization | Performance | Measure, analyze, and improve with evidence |
| Documentation | Core | Write clear, accurate, useful documentation |
| Git Workflow | Core | Professional version control practices |
| Docker & Containers | DevOps | Build and optimize containerized applications |
| CI/CD Pipeline Design | DevOps | Automate testing, building, and deployment |
| Database Design | Backend | Design, optimize, and review schemas |
| API Design | Backend | Build robust, consistent, developer-friendly APIs |
| Frontend Engineering | Frontend | Responsive, accessible, performant UIs |
| Backend Engineering | Backend | Reliable, scalable, secure server-side systems |
| UI/UX Pro Max | Frontend Design | Advanced UI/UX design intelligence, palettes, typography, accessibility, charts, and design systems |
| Design Resources Library | Frontend Design | Curated design systems, component libraries, and UI resources |
| Scrollytelling | Frontend Engineering | Engineer scroll-driven interactive websites with GSAP, WebGL, and smooth scroll |
TypeScript Libraries & Frameworks
| Skill | Category | Purpose |
|---|---|---|
| TypeScript | Language | TypeScript strict typing, generics, and advanced patterns |
| Vite | Tooling | Fast frontend builds and development server |
| React Styling | Frontend | Styling strategies for React applications |
| Ant Design | Frontend | Enterprise React UI components and design system |
| TanStack Query | Frontend | Server-state data fetching, caching, and synchronization |
| Redux Toolkit | Frontend | Predictable global state management for React |
| NestJS | Backend | Scalable, enterprise-ready Node.js server framework |
| Prisma | Backend | Type-safe ORM and database migrations |
| tRPC | Backend | End-to-end typesafe APIs without schema duplication |
| Zod | TypeScript | Schema validation with static type inference |
| Deno | Runtime | Secure, modern TypeScript/JavaScript runtime |
| n8n | Automation | Workflow automation and service integration |
iOS & Apple Platform
| Skill | Category | Purpose |
|---|---|---|
| iOS Development | Mobile | Build production iOS apps with Swift 6, SwiftUI, TCA, and HIG compliance |
| iOS NFC ID Reader | Mobile / NFC | Read Iraqi national ID and passports via CoreNFC with ICAO 9303 |
| iOS Security | Mobile / Security | Keychain, biometric auth, jailbreak detection, encryption, certificate pinning |
AI & Operations
| Skill | Category | Purpose |
|---|---|---|
| MCP Builder | AI | Build Model Context Protocol servers |
| AI Engineering | AI | Build reliable systems leveraging AI |
| Knowledge Distillation | Meta | Extract reusable lessons from projects |
| Quality Audit | Meta | Rigorously evaluate assets before acceptance |
| Incident Response | Ops | Systematic response to production incidents |
| Root Cause Analysis | Core | Determine fundamental causes, not symptoms |
| Stack Update Preflight | Meta | Inspect and update dependencies before every build task |
Elite & Advanced Capabilities
| Skill | Category | Purpose |
|---|---|---|
| 3D Scene Composition | 3D | Compose complex 3D scenes from reusable components in React |
| Advanced Cognitive Architecture | AI | Multi-step planning, tree-of-thought, and structured agent workflows |
| Agent Self-Healing | Security | Detect and repair vulnerabilities or errors in the agent's own code |
| AI-Driven Design System | Design | Generate and apply complete design systems using AI |
| Contextual UI Adaptation | Frontend | Adapt UIs based on device, viewport, and user preferences |
| Knowledge Graph Integration | AI | Build and query knowledge graphs automatically |
| Secure Execution Environment | Security | Run agent tasks in isolated, sandboxed environments |
| Semantic Memory Fabric | AI | Store, retrieve, and evolve contextual knowledge using vector search |
| Threat Intelligence Integration | Security | Monitor threat feeds and update security rules automatically |
| Visual Feedback Loop | Frontend | Analyze UI screenshots to detect visual issues and propose improvements |
| Urban Grid 3D Sim | Simulation | Interactive 3D urban electricity grid simulation |
Pmndrs Ecosystem
| Skill | Category | Purpose |
|---|---|---|
| React Three Fiber | Frontend Engineering | Build 3D scenes in React with @react-three/fiber |
| React Three Drei | Frontend Engineering | Helpers and abstractions for @react-three/fiber |
| React Three A11y | Frontend Engineering | Accessibility for React Three Fiber |
| React Postprocessing | Frontend Engineering | Postprocessing effects for React Three Fiber |
| React Spring | Frontend Engineering | Spring physics animation for React |
| Zustand | Frontend Engineering | Bearbones state management for React |
| Jotai | Frontend Engineering | Primitive and flexible state management |
| Valtio | Frontend Engineering | Proxy state management for React and Vanilla |
| UI Kit | Frontend Engineering | UI components for React Three Fiber |
| XR | Frontend Engineering | VR/AR for React Three Fiber |
| Leva | Frontend Engineering | GUI controls for React |
| Prai | Frontend Engineering | AI helpers from pmndrs |
| Viverse | Frontend Engineering | Web-based metaverse tools |
| Docs | Documentation | pmndrs documentation infrastructure |
22 Immutable Rules
Located in .agents/rules/, each rule references the Constitution article it derives from and includes a verification checklist.
| # | Rule | Focus |
|---|---|---|
| 1 | Engineering Principles | Never assume, reuse, keep simple |
| 2 | Architecture | SOLID, composition, loose coupling |
| 3 | Security | OWASP, input validation, secrets |
| 4 | Performance | Profile first, optimize hot paths |
| 5 | Debugging | Reproduce, isolate, root cause |
| 6 | Planning | Understand before solving, small steps |
| 7 | Documentation | Update README, explain decisions |
| 8 | Git | Small commits, meaningful messages |
| 9 | Testing | TDD, deterministic, edge cases |
| 10 | Code Review | Correctness first, constructive |
| 11 | Communication | Concise, specific, evidence-based |
| 12 | UX | Responsive, accessible, Core Web Vitals |
| 13 | API Design | Consumer-first, versioning, errors |
| 14 | Database | Parameterized queries, migrations |
| 15 | AI Engineering | Verify output, structured outputs |
| 16 | MCP | Single purpose, validate, document |
| 17 | Memory | Generalize lessons, update skills |
| 18 | Learning | Extract knowledge, create assets |
| 19 | Deployment | Automated, blue-green, rollback |
| 20 | Monitoring | Structured logs, actionable alerts |
| 21 | Quality | 7-gate audit, scoring rubric, reviewer ethics |
| 22 | Continuous Learning | Extract lessons and update the knowledge base after every task |
Self-Improving Knowledge Engine
DevinOS is not static. It is designed to learn and grow from every project it touches.
How It Works
Every Devin Project
|
v
Knowledge Distillation
|
v
Generalized Lessons
|
v
New Skills / Updated Rules
|
v
Pull Request to DevinOS
|
v
Community Review & Merge
|
v
Every Devin User Benefits
For Individual Users (Private Fork Model)
- Fork DevinOS into your own private repository
- Use it across all your projects
- After each project, Devin asks: "What did I learn? Can this become a reusable skill?"
- Extract lessons, generalize them, and add to your private fork
- Optionally contribute back to the public DevinOS (opt-in)
For the Community (Federated Learning)
- Weekly community updates aggregate validated learnings
- High-quality contributions from real projects are merged into the main branch
- Every user who pulls the latest DevinOS gets the collective wisdom of thousands of projects
- No code is ever shared automatically — only generalized, anonymized patterns
Privacy First
DevinOS never auto-uploads your code. The knowledge distillation process:
- Removes all project-specific identifiers
- Generalizes patterns into reusable skills
- Requires explicit opt-in before any contribution
- Keeps your private fork completely under your control
Repository Structure
DevinOS/
├── ENGINEERING_CONSTITUTION.md # Supreme law of DevinOS
├── README.md # This file
├── ROADMAP.md # Future plans
├── CHANGELOG.md # Version history
├── CONTRIBUTING.md # How to contribute
├── CODE_OF_CONDUCT.md # Community standards
├── SECURITY.md # Security policy
├── .gitignore # Cross-platform ignore rules
│
├── .agents/
│ ├── skills/ # 40 skills
│ ├── rules/ # 21 foundational rules
│ ├── workflows/ # Reusable workflows
│ ├── playbooks/ # Step-by-step guides
│ ├── memory/ # Knowledge base
│ ├── templates/ # Implementation templates
│ └── prompts/ # Curated prompts
│
├── docs/ # Documentation
├── examples/ # Example implementations
└── assets/ # Images and media
Quality Gates
Every skill must pass these gates before acceptance:
- Uniqueness - No duplicates in the repository
- Documentation - Description, purpose, trigger, context, workflow
- Examples - Good and bad examples included
- Anti-patterns - Common mistakes documented
- Verification - Checklist for correctness
- References - Further reading provided
- Framework Agnostic - General solution unless specific is unavoidable
- Production Quality - No shortcuts, no TODOs without tickets
Contributing
See CONTRIBUTING.md for the full guide.
Quick start:
- Read the Engineering Constitution
- Check ROADMAP.md for planned features
- Search existing issues to avoid duplicates
- Follow the skill template structure
- Submit a pull request with clear justification
Roadmap
| Version | Focus | Status |
|---|---|---|
| v0.1 | Foundation - Constitution, Rules, Skills | Live |
| v0.2 | Rules & Governance expansion | Planned |
| v0.3 | Skills expansion (50+) | Planned |
| v0.4 | Workflows library | Planned |
| v0.5 | Playbooks (Incident, Security, Optimization) | Planned |
| v0.6 | Memory system | Planned |
| v0.7 | Templates library | Planned |
| v0.8 | Prompt library (100+) | Planned |
| v0.9 | Knowledge graph & searchable index | Planned |
| v1.0 | Public release with community marketplace | Planned |
Full details: ROADMAP.md
Keywords
DevinOS, Devin AI, AI coding agent, engineering operating system, software engineering rules, AI prompt engineering, MCP server, Model Context Protocol, Devin skills, Devin rules, AI software development, automated code review, AI debugging, knowledge distillation, self-improving AI, LLM engineering, AI best practices, software architecture, secure coding, CI/CD automation, Docker best practices, frontend engineering, backend engineering, database design, API design, incident response, root cause analysis, AI engineering, open source AI tools, AI agent framework.
License
MIT - Free for personal and commercial use.
No comments yet
Be the first to share your take.