Skip to content

MASTER_PROMPT.md

Latest commit

 

History

History
118 lines (92 loc) · 3.87 KB

MASTER_PROMPT.md

File metadata and controls

118 lines (92 loc) · 3.87 KB

Master Prompt for Cursor - Sync (Expert Edition)

You are building Sync, a couple's shared chat MVP. Your task is to implement the product step-by-step, milestone by milestone.

⚠️ RULE: After completing each milestone, you must STOP, show me a checklist of what you did, provide code/tests/docs/demos, and explicitly ASK for my approval before starting the next milestone. Never skip or merge milestones.

🎯 Core MVP Scope

  • Shared couple chat with AI reflection → clarification → optional micro-actions
  • Strict safety boundaries, EU resources only, delete-only retention
  • Turn-taking enforcement (no simultaneous typing)
  • Mobile-first with optional web demo
  • Production-ready with proper security, testing, and deployment

🏗️ Architecture

  • Mobile App: React Native (Expo)
  • Website: Next.js (App Router)
  • API: Node.js + TypeScript (Fastify) + PostgreSQL
  • AI: OpenAI with custom orchestrator
  • Security: JWT, CORS, RLS, AES-GCM encryption

📁 Monorepo Structure

/app-mobile/     # React Native (Expo)
/website/        # Next.js site + /demo trial chat
/services/api/   # Fastify API + OpenAPI YAML
/services/ai/    # Orchestrator, prompts, evals
/packages/ui/    # Shared UI tokens/components
/packages/types/ # Shared TS types + generated client
/infra/          # DB migrations, env templates, KMS helpers
/docs/           # RFCs, runbooks, openapi.yaml

🚀 Development Milestones

M1: Foundation (API + Auth + Basic Chat)

  • PostgreSQL setup with RLS
  • JWT authentication
  • Basic message CRUD
  • Simple AI integration
  • Deliverable: Working API with auth

M2: Turn-Taking + Long-Polling

  • Session state machine
  • Real-time message sync
  • Turn-taking enforcement
  • Deliverable: Two-device chat flow

M3: AI Orchestrator + Safety

  • Custom AI prompt engineering
  • Safety boundary detection
  • Response validation
  • Deliverable: AI-powered conversation flow

M4: Safety & Boundary System

  • Tier-1 content filtering
  • Safety templates
  • EU support resources
  • Frontend safety locks
  • Deliverable: Production-ready safety system

M5: Survey + Delete

  • 3-emoji feedback system
  • Hard delete functionality
  • Survey analytics
  • Deliverable: User feedback and data deletion

M6: Website (Next.js)

  • Demo chat interface
  • Staging gates
  • Theme system
  • Deliverable: Web demo with staging protection

M7: Polish

  • Accessibility (WCAG AA)
  • Rate limiting
  • Copy review
  • Comprehensive testing
  • Deliverable: Production-ready application

🔒 Security Requirements

  • Encryption at rest: AES-GCM on sensitive columns
  • EU-only resources: All data processing in EU region
  • Delete-only retention: No data export, only deletion
  • Safety boundaries: AI content filtering and boundary detection
  • Turn-taking enforcement: No simultaneous typing

🛠️ Technical Stack

  • Backend: Node.js, TypeScript, Fastify, PostgreSQL
  • Frontend: React Native (Expo), Next.js
  • AI: OpenAI GPT-4 with custom orchestrator
  • Security: JWT, CORS, RLS, AES-GCM
  • Deployment: Vercel (web), Expo (mobile)

📋 Development Process

  1. Milestone-by-milestone: Never skip or merge milestones
  2. Explicit approval: Ask for approval before proceeding
  3. Comprehensive deliverables: Code, tests, docs, demos
  4. Production-ready: Each milestone must be deployable
  5. Security-first: All features must meet security requirements

🎯 Success Criteria

  • Functional: Complete couple chat with AI assistance
  • Secure: Production-ready security and privacy
  • Scalable: Proper architecture for growth
  • Maintainable: Clean code with comprehensive tests
  • Accessible: WCAG AA compliance
  • Deployable: Ready for production deployment

Remember: Always ask for explicit approval before starting the next milestone. Never skip or merge milestones.