🚀 Vibe-Coding Workflow: AI-Powered MVP Development

Build an MVP in hours, not months – guided end-to-end by AI.

---

🎯 What This Does

Turn any app idea into working code through 4 AI-powered stages:

#	Stage	Goal	Output
1️⃣	Research	Validate market & tech landscape	Research findings
2️⃣	Define	Clarify product scope	PRD (Product Requirements)
3️⃣	Design	Decide how to build	Technical Design doc
4️⃣	Build	Generate & test code	Working MVP

---

🏃 Quick Start

The Entire Workflow in 30 Seconds

Step	What You Do	Time	Result
📚	Copy prompts → Answer questions	20 min	Research doc
📝	Define your app idea	15 min	PRD doc
🏗️	Choose technical approach	15 min	Tech Design doc
🤖	Generate AI instructions	10 min	NOTES.md
💻	Tell AI: "Read NOTES.md and build"	1-3 hrs	Working MVP!

That's it!

---

✅ Prerequisites

🤖 AI Platform (choose one)

• AI Studio ⭐ - Best free limits (10 req/min)
• Claude - (~40 msg/day free)
• ChatGPT - (10 msg/per 5hr)
• Gemini - (limited access to 2.5 Pro for free users | 100 msg/day for paid)

💻 AI-Enabled IDE / Agent (choose one)

• Jules - Asynchronous agent that works on tasks independently.
• Claude Code ⭐ - Powerful command-line tool for agentic coding.
• Cursor ⭐ - Purpose-built for AI coding.
• VS Code + Claude Extension
• VS Code + GitHub Copilot - Popular combo.
• Windsurf - Privacy-focused.
• Cline - Open source VS Code extension.

🛠 Basic Requirements

• Any modern browser
• 2-4 hours of time
• Zero coding experience needed!

---

📋 The 4-Step Workflow

Transform your idea into a working MVP through these guided steps:

1️⃣ Deep Research 🔍

Validate your idea with AI-powered market research • 20-30 min • Creates research-*.txt

What this does: Analyzes market opportunity, competitors, and technical feasibility.

How it works:

1. Copy the entire part1-deepresearch.md file.
2. Paste into any AI platform (AI Studio, Claude, ChatGPT, etc.).
3. Answer 5-6 questions about your idea (tailored to your experience level).
4. AI generates comprehensive research on market, competitors, and implementation options.
5. Save the output as research-[YourAppName].txt.

2️⃣ Product Requirements (PRD) 📝

Define exactly what you're building • 15-20 min • Creates PRD-*.md

What this does: Transforms your idea into clear, actionable product specifications.

How it works:

1. Copy part2-prd-generator.md into a new Gemini chat.
2. Attach your research findings when prompted.
3. Answer questions about features, users, and success metrics.
4. AI creates a professional PRD document.
5. Save as PRD-[YourAppName]-MVP.md.

3️⃣ Technical Design 🏗️

Plan the technical architecture • 15-20 min • Creates TechDesign-*.md

What this does: Decides the tech stack and implementation approach.

How it works:

1. Copy part3-tech-design-generator.md into a new Gemini chat.
2. Attach your PRD (required) and research (optional).
3. Answer questions about platform, complexity, and constraints.
4. AI designs the technical architecture.
5. Save as TechDesign-[YourAppName]-MVP.md.

4️⃣ Generate AI Agent Instructions 🤖

Create a blueprint for your AI coding assistant • 5-10 min • Creates NOTES.md

What this does: Converts all previous docs into step-by-step coding instructions.

How it works:

1. Copy part4-generate-notes-for-agent.md into a new Gemini chat.
2. Attach both PRD and Technical Design documents.
3. AI creates detailed implementation instructions.
4. Save the output as NOTES.md.

5️⃣ Build with AI IDE 💻

Time 1-3 hrs • Tool Your AI IDE • Output Working MVP

➊ Project Setup

1. Create a project folder → add a docs subfolder.
2. Move generated files:

   docs/research-*.txt
   docs/PRD-*.md
   docs/TechDesign-*.md
   NOTES.md        (in root, not docs)
   

3. Create an agent instructions file in your root folder (see step ➋).
4. Open the folder in your AI-enabled IDE or connect it to your AI agent.

➋ Create AI Agent Rules

The filename and setup depend on your chosen tool. Create the appropriate file in your project's root directory.

Tool	Rule File	Setup
Jules & Codex	AGENTS.md	Create this file to describe your project's tools, conventions, and goals. Jules reads this automatically.
Claude Code (CLI)	CLAUDE.md	Create this file to provide context, commands, and style guides. The Claude CLI reads this automatically.
Cursor	.cursorrules	Create this file and paste the rule content below.
GitHub Copilot	copilot-instructions.md	Create this file and paste the rule content below.

General Rule Content (for Cursor, Copilot, etc.):

Read NOTES.md first – it is your implementation guide.
Build features incrementally and test each one.

# Code Philosophy:
- Prioritize simplicity: less code is more maintainable
- Replace rather than patch: rewrite components instead of adding layers
- Fail fast: no fallback mechanisms that mask errors
- Keep it clean: flag deprecated files for removal
- Stay modular: try to have 500 lines per file, keep the structure modular

# Development Standards:
- Thread safety: prevent race conditions through proper state management
- Configuration: use environment variables, never hardcode secrets or URLs
- Documentation: functions need a clear docstring explaining its purpose
- Structure: organize code into logical modules with single responsibilities

# Output Requirements:
- Show complete implementations (no "rest remains the same" comments)
- Provide exact placement instructions for code snippets
- Include full context for single function changes
- Display entire modified files unless explicitly asked for snippets

# AI Behavior:
- Use extended thinking mode for complex logic – thorough analysis prevents bugs
- Explain architectural decisions before implementing
- Validate against requirements before generating code

# IMPORTANT:
- Create a [project-name].md file if you haven't.
- After adding a major feature or completing a milestone, update [project-name].md.
- Always read [project-name].md before writing any code.
- Document the entire database schema in [project-name].md.
- For new migrations, make sure to add them to the same file.

💡 Agent-Specific Setup

• For Jules: Your AGENTS.md should describe the project's purpose and any key architectural patterns or commands.
• For Claude Code: Your CLAUDE.md should contain instructions, common commands, and style guides to improve accuracy. You can run /init in the Claude Code CLI to generate a template.

➌ Kick-off Prompt (pick one)

Level	Prompt
Vibe-Coder	"I'm a non-technical founder. Read NOTES.md, explain the project, then guide me step-by-step."
Learner	"Read NOTES.md, explain the tech stack, teach me as we build."
Developer	"Read NOTES.md and the files in the /docs folder. Confirm the architecture, then start implementing Phase 1."

➍ Follow-Up Prompts

Planning & Implementation

• "Show me the implementation plan for [Feature] before we code"
• "Let's build [Feature] - start with the simplest working version"
• "What are the edge cases we should handle for [Feature]?"

Debugging & Understanding

• "I'm seeing [error/issue] - explain what's wrong and how to fix it"
• "Break down this function and explain each part's purpose"
• "Why did you choose this approach over alternatives?"

Testing & Validation

• "Create a test to verify [Feature] works correctly"
• "Let's manually test the happy path for [Feature]"
• "Show me what could break this feature"

Progress Tracking

• "What have we built so far vs. what's in NOTES.md?"
• "What's the next priority from our requirements?"
• "Are we still aligned with the original PRD?"

➎ Launch Checklist

Before calling it done:

• ✅ Documentation - "Generate a user-friendly README with quick start steps"
• ✅ Configuration - "Create .env.example with all needed settings and descriptions"
• ✅ Validation - "Let's do a complete walkthrough as a new user would"
• ✅ Error Handling - "Show me how the app handles common failure scenarios"
• ✅ Deployment - "Generate deployment instructions based on our Tech Design"

🎉 Ship it! Your MVP is ready for real users.

---

📁 Project Structure (after Step 5)

your-app/
├─ docs/
│  ├─ research-YourApp.txt
│  ├─ PRD-YourApp-MVP.md
│  └─ TechDesign-YourApp-MVP.md
├─ .clauderules    ← or .cursorrules, etc. (based on your IDE)
├─ NOTES.md        ← AI implementation guide
├─ README.md       ← setup & run instructions (AI-generated)
├─ .env.example    ← environment variables template
├─ src/…           ← your app code (AI-generated)
└─ …

---

💡 Pro Tips

• Be Specific: The more detailed your request, the better the result. Include examples when possible.
• Plan First: Ask for an implementation plan before coding – prevents costly rewrites.
• Test Early: Request tests before features – clarifies requirements and catches bugs.
• Stay Focused: Build only the core MVP features from your PRD. Enhancement comes later.
• Verify Often: Run the app after each major addition – catch issues immediately.
• Learn as You Go: Ask "why" whenever confused – AI loves to teach.

Common Pitfalls

❌ Pitfall	✅ Fix
Skipping research	Complete Step 1 – it shapes everything
Feature creep	Stick to the PRD MVP scope
Perfectionism	Ship working code, polish later
Ignoring errors	Address issues immediately, don't accumulate technical debt

---

🚨 Troubleshooting

Common Issues & Quick Fixes

Problem	Solution
"AI ignores my documents"	Say: "First, read the attached [document name], then proceed"
"Lost track of progress"	Ask: "Compare our current state against NOTES.md - what's done and what's next?"
"Features don't match PRD"	Tell AI: "This implementation differs from the PRD. Review [feature] in the PRD and align the code"
"Code is getting messy"	Request: "Refactor this following our modular structure rules - max 500 lines per file"
"AI is overcomplicating"	Remind: "Simplify this - remember, less code is better for our MVP"

Still Stuck?

• Start a fresh chat with all documents attached
• Lead with: "I'm building an MVP. Read all attached docs, confirm understanding, then proceed with [specific task]"

---

🤝 Contributing

PRs & issues welcome!

• 🐛 Report issues with prompts
• 💡 Share your success stories
• 🔧 Improve templates for edge cases
• 📚 Add examples of built MVPs

---

📜 License

Released under the MIT License.

---

The goal isn't perfect code on the first try – it's a working MVP you can iterate on.
Every great app starts somewhere. Yours starts now. 🚀