Production-grade rules, docs, and patterns for AI-native full-stack development. Works with Cursor and Claude Code.
15 engineering rules + 9 doc templates + skills guide + agent patterns + MCP setup. Drop into any project — your AI coding tool immediately thinks like a senior engineer.
curl -fsSL https://raw.githubusercontent.com/aiagentwithdhruv/ai-coding-rules/main/install.sh | bashcurl -fsSL https://raw.githubusercontent.com/aiagentwithdhruv/ai-coding-rules/main/claude/CLAUDE.md -o CLAUDE.mdcurl -fsSL https://raw.githubusercontent.com/aiagentwithdhruv/ai-coding-rules/main/install.sh | bash
curl -fsSL https://raw.githubusercontent.com/aiagentwithdhruv/ai-coding-rules/main/claude/CLAUDE.md -o CLAUDE.mdai-coding-rules/
├── rules/ # 15 Cursor rules (.mdc)
├── claude/ # Claude Code setup
│ ├── CLAUDE.md # All 15 rules in one file
│ ├── compose.sh # Build custom CLAUDE.md
│ ├── rules/ # 15 individual rule files
│ └── .claude/settings.json # Permission config template
├── docs/ # 9 project templates
│ ├── PRD.md # Product requirements
│ ├── ARCHITECTURE.md # System architecture + data flow
│ ├── API_SPEC.md # API endpoints + response format
│ ├── DB_SCHEMA.md # Tables, columns, indexes
│ ├── DEPLOYMENT.md # Local + AWS + Vercel + VPS + CI/CD
│ ├── SKILLS.md # How to create and use skills
│ ├── AGENTS.md # Subagent patterns and templates
│ ├── LOADOUT.md # Project manifest template
│ └── MCP.md # MCP server setup guide
├── install.sh # One-liner installer
└── LICENSE
| # | Rule | What it enforces |
|---|---|---|
| 00 | Global Architect | Think architect-first. Clean architecture. Separation of concerns. |
| 10 | Backend FastAPI | Thin routes. Services for logic. Repositories for DB. |
| 20 | Frontend Next.js | TypeScript. Small components. Centralized API clients. |
| 30 | Database PostgreSQL | Migrations. Indexes. Foreign keys. No raw SQL in routes. |
| 35 | API Contracts | Version APIs. Typed schemas. Consistent error format. |
| 40 | Cache Redis | Intentional TTLs. Stable keys. Wrapped access. |
| 45 | Environment Config | Validate at startup. .env.example. Secret managers. |
| 50 | RAG System | Separate ingestion from generation. Chunk metadata. |
| 55 | Data & Model Versioning | Version datasets. Named checkpoints. Reproducibility. |
| 60 | Agents | Tool schemas. Validated outputs. Supervisor patterns. |
| 70 | Security | No hardcoded secrets. Prompt injection resistance. |
| 80 | Testing & Quality | Tests for critical paths. Deterministic units. Linting. |
| 85 | Error & Observability | Structured errors. Request tracing. Health checks. |
| 90 | DevOps & Deployment | Docker. AWS. Vercel. VPS. CI/CD. |
| 99 | Response Style | Minimal changes. Production code. No toy implementations. |
| Doc | What it defines | Why it matters |
|---|---|---|
| PRD.md | Product requirements, users, milestones | AI knows what to build |
| ARCHITECTURE.md | Layers, data flow, key decisions | AI knows how it's structured |
| API_SPEC.md | Endpoints, response format, patterns | AI writes consistent APIs |
| DB_SCHEMA.md | Tables, columns, types, indexes | AI writes correct queries |
| DEPLOYMENT.md | Local + cloud + CI/CD setup | AI writes deploy-ready code |
| SKILLS.md | Reusable automation capabilities | AI reuses before rebuilding |
| AGENTS.md | Subagent patterns, templates | AI delegates complex tasks |
| LOADOUT.md | Project manifest template | AI gets instant context |
| MCP.md | External tool integrations | AI connects to services |
"Use clean architecture"
"Never hardcode secrets"
"Use Pydantic for validation"
"We have 12 tables with multi-tenant isolation"
"API uses /api/v1/ prefix with JWT auth"
"Deploy to AWS ECS with GitHub Actions"
"Use the deploy skill to push to production"
"Use the scrape-leads skill for lead generation"
The 00-global-architect rule auto-references your docs:
Before implementing, align with docs/PRD.md, docs/ARCHITECTURE.md,
docs/API_SPEC.md, docs/DB_SCHEMA.md, and docs/DEPLOYMENT.md if present.
| Cursor | Claude Code | |
|---|---|---|
| Format | .mdc with YAML frontmatter |
.md plain markdown |
| Location | .cursor/rules/ |
CLAUDE.md at project root |
| Multiple files | 15 numbered files | 1 combined or subdirectory split |
| Triggers | alwaysApply, globs, manual | Always loaded from current dir |
| Install | install.sh |
Single file copy |
Use both in the same project — zero conflict:
your-project/
├── .cursor/rules/*.mdc # Cursor reads these
├── CLAUDE.md # Claude Code reads this
└── docs/ # Both reference these
Don't use Next.js? Replace 20-frontend-nextjs:
- React + Vite, SvelteKit, Nuxt, Angular
Don't use FastAPI? Replace 10-backend-fastapi:
- Express, Django, Go, Rails
Cursor — create 05-my-rule.mdc in .cursor/rules/ (gap numbering, fits between 00 and 10).
Claude Code — add a section to CLAUDE.md.
Pick only what you need:
cd ai-coding-rules/claude
./compose.sh 00 10 30 70 80 99 > ~/project/CLAUDE.md # Backend only
./compose.sh 00 20 70 80 99 > ~/project/CLAUDE.md # Frontend only
./compose.sh 00 50 55 60 70 99 > ~/project/CLAUDE.md # AI/ML only
./compose.sh > ~/project/CLAUDE.md # Everything- skills — 38 reusable AI agent skills (lead gen, email, video, deployment). Install:
npx skills add aiagentwithdhruv/skills
ai-coding-rules = how to set up your AI coding environment
skills = what to automate with AI agents
Built by @AiwithDhruv with Claude Code.
Production-tested across 30+ projects — AI agents, RAG systems, SaaS products, and client work.
MIT — use it, fork it, ship better code.