Skip to content

lvlrf/MegaPromp-AI-Engeering

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

18 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

MegaPrompt V8 - AI-Led Vibe Coding Framework

Transform ideas into production-ready applications through structured AI-led development

Version License AI Models


🎯 What is MegaPrompt V8?

MegaPrompt V8 is a comprehensive framework that enables AI models (Claude, GPT, Gemini) to transform natural language descriptions into complete, production-ready applications through structured phases and intelligent execution.

Key Philosophy:

  • 🧠 AI-Led Execution: AI drives technical implementation
  • πŸ‘€ User-Led Vision: Users define goals and requirements
  • 🀝 Collaborative Partnership: Continuous dialogue and iteration
  • πŸ“Š Evidence-Based Decisions: All recommendations backed by data
  • ⚑ Proactive Intelligence: AI suggests next steps autonomously

✨ Features

Core Capabilities

  • βœ… Phase-Based Development: Structured flow from discovery to delivery
  • βœ… Multi-Model Support: Works with Claude, GPT, Gemini
  • βœ… Comprehensive Documentation: 30+ files auto-generated
  • βœ… Cost Optimization: Real-time tracking and multi-model strategies
  • βœ… Security First: Automated security checks before delivery
  • βœ… Quality Gates: 80%+ test coverage, automated validation
  • βœ… Technology Radar: Continuous evaluation of new technologies
  • βœ… Scalability Planning: Plan for scale, build essentials now

Advanced Features

  • πŸ”„ Dynamic Role Switching: 30+ expert roles (DevOps, Security, ML, etc.)
  • 🎨 Flexible UI/UX Modes: Support for templates, custom designs, AI generation
  • πŸ’Ύ Context Memory: Persistent decision tracking across sessions
  • 🚨 Telegram Notifications: Real-time progress alerts
  • πŸ”” Sound Alerts: Audio notifications for key events
  • πŸ“¦ Git Auto-Backup: Automated version control every 5 tasks
  • πŸ€– Multi-Agent Coordination: Parallel execution with multiple AI models

πŸš€ Quick Start

For AI Models

1. Identify Your Model

β–‘ Claude β†’ Read CLAUDE.md
β–‘ GPT β†’ Read CODEX.md
β–‘ Gemini β†’ Read GEMINI.md

2. Load Framework

Read: MEGAPROMPT_V8_FINAL.md (complete instructions)
Read: AGENT_START_PROMPT.md (quick setup)

3. Start Working

Begin with Phase 0 greeting
Follow structured workflow
Generate comprehensive documentation

For Users

1. Choose Your AI

  • Claude Sonnet 4.5 (balanced, fast)
  • Claude Opus 4.1 (strongest reasoning)
  • GPT-4o (UI/UX excellence)
  • Gemini 2.0 Flash (speed, efficiency)

2. Start Conversation

User: "I want to build [your idea]"
AI: [Follows MegaPrompt V8 workflow]

3. Collaborate

  • AI asks questions (Discovery)
  • You provide answers
  • AI suggests options
  • You make decisions
  • AI builds everything

4. Receive Deliverable

  • Complete documentation
  • Production-ready code
  • Tests (80%+ coverage)
  • README and deployment guide
  • ZIP package ready to use

πŸ“‚ File Structure

Agent Configuration Files

MEGAPROMPT_V8_FINAL.md         # Core framework (complete instructions)
AGENT_START_PROMPT.md          # Quick start guide
CLAUDE.md                      # Claude-specific instructions
CODEX.md                       # GPT-specific instructions
GEMINI.md                      # Gemini-specific instructions
.cursorrules                   # Cursor IDE integration
README.md                      # This file

Generated Project Structure

docs/
  00_context/                  # Requirements & decisions
    PRD.md                     # User's vision (their words)
    PRD_ai.md                  # Executable plan
    PRD_NOTES.md               # Execution log
    PRD_IMPLEMENTATION_MATRIX.md  # Feature→Task→File mapping (if >20 tasks)
    DECISIONS.md               # All technical decisions
    GLOSSARY.md                # Domain terminology
    CONTEXT_MEMORY.md          # Persistent memory

  10_product/                  # Features & planning
    SPEC.md                    # Detailed specifications
    TASKS.md                   # Task breakdown
    MVP_CHECKLIST.md           # Scope control

  20_engineering/              # Architecture & tech
    ARCHITECTURE.md            # System design
    RULES.md                   # Coding standards
    TECH_STACK.md              # Technology rationale
    DATABASE_SCHEMA.md         # Data model
    SECURITY_CHECKLIST.md      # Security protocol

  30_design/                   # UI/UX decisions
    UIUX.md                    # Design decisions
    UI_PAGES.md                # Page inventory
    UI_PROMPTS.md              # AI generation prompts
    UI_MODE.md                 # Design workflow

  40_api/                      # API specifications
    OPENAPI.yaml               # API spec (OpenAPI 3.0)
    OPENSPEC_WORKFLOW.md       # API workflow

  50_testing/                  # Testing strategy
    TEST_STRATEGY.md           # Testing approach

  90_ops/                      # Operations & tracking
    CHANGELOG.md               # Change history
    PROGRESS.md                # Real-time status
    COST_TRACKING.md           # Budget dashboard
    TIMELINE_AND_COST.md       # Planning
    DEPLOYMENT_GUIDE.md        # Deployment steps
    SCALE_PLAN.md              # Growth strategy
    TELEGRAM_SETUP.md          # Notification config
    SOUND_NOTIFICATION_SETUP.md   # Alert config
    GIT_AUTO_BACKUP.md         # Version control

src/                           # Source code
tests/                         # Test files
README.md                      # Project readme
.env.example                   # Environment template
package.json                   # Dependencies

🎨 Development Phases

Phase 0: START

AI greets user in their language
Explains the process
Asks what they want to build

Phase 1: DISCOVERY (Chat Mode)

AI asks strategic questions:
- What problem does this solve?
- Who are the users?
- What are the core features?
- Any constraints (time, budget, scale)?

AI extracts:
- User personas
- Use cases
- Data entities
- Scale requirements

Phase 2: PLANNING (Chat Mode)

AI proposes:
- Complete feature list (MVP vs Later vs Never)
- Technology options (with pros/cons/evidence)
- Architecture approach
- Timeline estimate
- Cost projection

User approves each decision
AI documents everything

Phase 3: BUILD (Execution Mode)

Triggered by: "BUILD" or "Ψ¨Ψ³Ψ§Ψ²"

AI generates:
- All documentation (30+ files)
- Complete source code
- Configuration files
- Tests (80%+ coverage)
- README and guides

AI updates PROGRESS.md after each task
AI suggests next steps proactively

Phase 4: DELIVERY

AI performs:
- Security scan (auto-check)
- Quality gates validation
- Test coverage verification
- Documentation completeness
- Final README generation

AI packages:
- ZIP file with everything
- Clear installation instructions
- Quick start guide

πŸ€– Multi-Model Strategy

For projects with 20+ tasks, MegaPrompt V8 can coordinate multiple AI models:

Recommended Distribution

High Complexity (Architecture, Security):
β†’ Claude Opus 4.1 / GPT o1

Medium Complexity (Backend, API):
β†’ Claude Sonnet 4.5 / GPT-4o

Low Complexity (Frontend, UI):
β†’ Gemini Flash 2.0 / GPT-4o

Testing (Fast iteration):
β†’ Gemini Flash 2.0

Documentation:
β†’ Gemini Pro 2.0 / GPT-4o

Cost & Time Comparison

Strategy Duration Cost Quality
Single Model 12h $25 β˜…β˜…β˜…β˜…β˜†
Multi-Model Balanced 7h $28 β˜…β˜…β˜…β˜…β˜…
Multi-Model Speed 5h $35 β˜…β˜…β˜…β˜…β˜†
Multi-Model Budget 10h $18 β˜…β˜…β˜…β˜†β˜†

πŸ’‘ Example Workflow

User Perspective

User: "I want to build a task management app"

AI: "Great! Let me understand your vision.
     
     What problem does this solve for users?
     Who will use it (personal, teams, both)?
     Any specific features you need?
     Timeline or budget constraints?"

User: [Answers questions]

AI: "Based on your input, here are all features:
     
     MVP (Build Now):
     ● User authentication
     ● Create/edit/delete tasks
     ● Mark tasks as complete
     
     Future (Later):
     β—‹ Team collaboration
     β—‹ File attachments
     β—‹ Email notifications
     
     Out of Scope:
     🚫 Mobile app (use PWA instead)
     
     Approve?"

User: "Yes"

AI: "Technology choices:
     
     Frontend: React vs Vue vs Svelte
     [Shows pros/cons/evidence for each]
     My recommendation: React
     Reason: [specific rationale]
     
     Your choice?"

User: "React"

AI: "Perfect! Ready to build?
     
     Timeline: 2 weeks
     Cost: $25 (using Claude Sonnet)
     Deliverable: Complete app + docs
     
     Say 'BUILD' when ready"

User: "BUILD"

AI: [Generates all files, provides progress updates]

AI: "πŸ“¦ Project complete!
     
     [Download ZIP - 2.3MB]
     
     Quick start:
     1. Extract ZIP
     2. npm install
     3. npm run dev
     
     Enjoy! πŸš€"

πŸ”’ Security Features

Automated Security Checks

Before every delivery:

βœ… API keys in .env (not hardcoded)
βœ… Passwords hashed (bcrypt/argon2)
βœ… SQL injection prevention
βœ… CORS properly configured
βœ… Rate limiting implemented
βœ… Dependencies scanned for vulnerabilities

Security Checklist File

Generated: docs/20_engineering/SECURITY_CHECKLIST.md

Contains:

  • Authentication best practices
  • Authorization patterns
  • Input validation rules
  • Error handling guidelines
  • Dependency security
  • Deployment security

πŸ“Š Cost Tracking

Real-Time Dashboard

File: docs/90_ops/COST_TRACKING.md

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ Cost Dashboard                              β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ Current:  $18.50 / $25.00 (74%)             β”‚
β”‚ Remaining: $6.50                            β”‚
β”‚ β–“β–“β–“β–“β–“β–“β–“β–‘β–‘β–‘ 74%                              β”‚
β”‚                                             β”‚
β”‚ Projection: $24.00 (On track βœ“)            β”‚
β”‚ Buffer: $1.00                               β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

Per Model:
- Claude Sonnet: $8.70
- GPT-4o: $6.50
- Gemini Flash: $3.30

Status: βœ“ On Track

🎯 Quality Standards

Testing Requirements

  • Minimum Coverage: 80%
  • Unit Tests: All critical functions
  • Integration Tests: API endpoints
  • E2E Tests: Core user flows

Code Quality

  • TypeScript: Type safety enforced
  • ESLint: No warnings allowed
  • Prettier: Consistent formatting
  • Documentation: JSDoc for all functions

Delivery Checklist

β–‘ All tasks completed (100%)
β–‘ Tests passing (80%+ coverage)
β–‘ Security scan clean (no critical issues)
β–‘ Dependencies updated (no vulnerabilities)
β–‘ Documentation complete (all files)
β–‘ README generated (installation guide)
β–‘ .env.example provided (configuration)
β–‘ Deployment guide ready (step-by-step)

🌟 Advanced Features

Technology Radar

AI continuously evaluates new technologies and suggests when beneficial:

"πŸ’‘ New technology available: Bun 1.0

Benefits for this project:
- 3x faster than Node.js
- Built-in TypeScript support
- Drop-in replacement

Would you like to use it?
Migration effort: 30 minutes"

Context Memory

AI remembers decisions across sessions:

File: docs/00_context/CONTEXT_MEMORY.md

Previous decisions:
- Database: PostgreSQL (chosen over MongoDB)
- Auth: JWT (chosen over OAuth initially)
- Styling: Tailwind CSS (chosen over Material UI)

Reasoning preserved for future reference

Scalability Planning

AI plans for scale but builds essentials:

File: docs/90_ops/SCALE_PLAN.md

Now (MVP):
- Single server
- SQLite database
- Simple caching

1K-10K users:
- Load balancer
- PostgreSQL
- Redis cache

10K+ users:
- Kubernetes
- Database sharding
- CDN

πŸ”— Inspiration & Integration

MegaPrompt V8 draws inspiration from:

Cloudflare VibeSDK

GitHub: cloudflare/vibesdk

Learnings:

  • Phase-wise code generation
  • Container-based execution
  • Real-time preview capabilities
  • Durable Objects for state management

Cloudflare Agents

GitHub: cloudflare/agents

Learnings:

  • Stateful AI agent patterns
  • WebSocket communication
  • Persistent memory systems
  • Multi-agent coordination

πŸ“š Documentation

Core Files

  • MEGAPROMPT_V8_FINAL.md - Complete framework specification
  • AGENT_START_PROMPT.md - Quick start for AI agents
  • CLAUDE.md - Claude-specific instructions
  • CODEX.md - GPT-specific instructions
  • GEMINI.md - Gemini-specific instructions

Integration

  • .cursorrules - Cursor IDE configuration
  • README.md - This file (GitHub documentation)

🀝 Contributing

Want to improve MegaPrompt V8?

  1. Fork the repository
  2. Create your feature branch
  3. Make improvements
  4. Test thoroughly
  5. Submit pull request

Areas for contribution:

  • Additional AI model support
  • New documentation templates
  • Enhanced security checks
  • Performance optimizations
  • Language translations

πŸ“„ License

MIT License - See LICENSE for details


πŸ™ Acknowledgments

  • Cloudflare team for VibeSDK and Agents frameworks
  • AI research community for foundational work
  • Open source contributors

πŸ“ž Support


πŸš€ Get Started

  1. Choose your AI model (Claude/GPT/Gemini)
  2. Load the appropriate guide (CLAUDE.md/CODEX.md/GEMINI.md)
  3. Read MEGAPROMPT_V8_FINAL.md
  4. Start building!

Remember: AI leads execution, you lead vision. Together, you create amazing products.


Made with ❀️ using AI + Human Collaboration

Version 8.0 - Last Updated: 2024-12-31

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors