# vibe-coding-prompt-template
**Repository Path**: rwwang/vibe-coding-prompt-template
## Basic Information
- **Project Name**: vibe-coding-prompt-template
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No
## Statistics
- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-11-03
- **Last Updated**: 2025-11-03
## Categories & Tags
**Categories**: Uncategorized
**Tags**: None
## README
# π Vibe-Coding Workflow: AI-Powered MVP Development
[](LICENSE)
[](http://makeapullrequest.com)
[](README.md)
**Build an MVP in hours, not months β guided by the latest AI coding agents**
---
## π― What This Does
Transform any app idea into working code through 5 AI-powered stages using the latest 2025 AI models:
| # | Stage | Goal | Output |
|---|-------|------|--------|
| 1οΈβ£ | **Research** | Validate market & tech landscape | Research findings |
| 2οΈβ£ | **Define** | Clarify product scope | PRD (Product Requirements) |
| 3οΈβ£ | **Design** | Decide how to build | Technical Design doc |
| 4οΈβ£ | **Generate AI Agent Instructions** | Convert docs into agent blueprints | NOTES.md + agent configs |
| 5οΈβ£ | **Build** | Generate & test code | Working MVP |
---
## π Quick Start
β‘ The Entire Workflow in 60 Seconds
| Step | What You Do | Time | Result |
|:---:|-------------|:----:|--------|
| π | Copy prompts β Answer questions | 20 min | Research doc |
| π | Define your app idea | 15 min | PRD doc |
| ποΈ | Choose technical approach | 15 min | Tech Design doc |
| π€ | Generate AI instructions | 10 min | NOTES.md + agent config |
| π» | Tell AI: *"Read NOTES.md and build"* | 1-3 hrs | **Working MVP!** |
---
## β
Prerequisites
π€ AI Platform (Required - Choose One)
### Best Free Options
- **[AI Studio](https://aistudio.google.com)** β - Gemini 2.5 Pro free tier with the full 1,048,576-token context ([see model details](https://ai.google.dev/gemini-api/docs))
- **[Claude.ai](https://claude.ai)** - Claude Sonnet 4.5 with 200K default / 1M beta context ([see current versions](https://docs.anthropic.com))
- **[ChatGPT](https://chat.openai.com)** - GPT-5 via the Responses interface ([see OpenAI docs](https://developers.openai.com))
### Premium Options (Better for Complete Apps)
- **[Claude Pro](https://claude.ai)** - Sonnet 4.5 with higher rate limits and 1M-token projects ([pricing](https://claude.ai))
- **[Gemini Advanced](https://gemini.google.com)** - Gemini 2.5 Pro/Flash Ultra access with priority throughput ([pricing](https://gemini.google.com))
- **[ChatGPT Plus](https://chat.openai.com)** - GPT-5 and o-series reasoning models with higher allocations ([pricing](https://openai.com/chatgpt))
π» AI Coding Agent/IDE (Required - Choose One)
### Terminal-Based Agents (Advanced)
- **[Claude Code](https://github.com/anthropics/claude-code)** β - Project-aware CLI with session memory and automated testing ([docs](https://docs.anthropic.com))
```bash
npm install -g @anthropic-ai/claude-code
claude init # In your project directory
```
- **[Gemini CLI](https://github.com/google-gemini/gemini-cli)** - Free & open source with Gemini 2.5 streaming context ([docs](https://ai.google.dev/gemini-api/docs))
```bash
npm install -g @google-gemini/cli
gemini login # Connect your Google account
```
- **[OpenAI CLI](https://platform.openai.com/docs/guides/responses)** - Terminal access to GPT-5 and o-series tool-calling pipelines ([docs](https://platform.openai.com/docs))
```bash
pip install --upgrade openai
openai responses.create -m gpt-5 --input "Summarize NOTES.md"
```
### Async/Cloud Agents
- **[Jules by Google](https://jules.google/docs)** β - Vertex AI-connected async agent that executes scoped work packets
- **[GitHub Copilot Agent](https://github.com/features/copilot)** - Agentic PR, code review, and deployment prep automation ([docs](https://docs.github.com/en/copilot))
### IDE-Based Tools (Beginner Friendly)
- **[Cursor](https://cursor.sh)** β - AI editor with long-term Memory, Auto Debug, and Codebase Q&A ([pricing](https://cursor.sh))
- **[VS Code + Github Copilot](https://code.visualstudio.com/)** β - Editor with Copilot Agent Mode and inline pair programming ([Copilot pricing](https://github.com/features/copilot/plans))
### No-Code Platforms (Easiest)
- **[Bolt.new](https://bolt.new)** β - Instant Next.js/Supabase apps with scheduled automations ([pricing](https://bolt.new))
- **[Lovable](https://lovable.dev)** - AI fullstack builder shipping 25k apps daily ([pricing](https://lovable.dev))
- **[v0 by Vercel](https://v0.dev)** - AI UI composition with deployment-ready React components ([pricing](https://v0.dev))
π Basic Requirements
- Any modern browser
- 2-4 hours of time
- Basic computer skills (no coding required!)
- Optional: Node.js 20+ for terminal tools
---
## π The 5-Step Workflow
### 1οΈβ£ Deep Research π
Validate your idea with AI-powered market research β’ 20-30 min β’ Creates research-*.txt
**What this does:** Analyzes market opportunity, competitors, and technical feasibility using the latest AI models.
**How it works:**
1. Copy the entire `part1-deepresearch.md` file
2. Paste into AI Studio (for Gemini 2.5 Pro), Claude.ai (for Claude Sonnet 4.5), or ChatGPT (for GPT-5)
3. Answer 5-6 questions tailored to your experience level
4. AI generates comprehensive research with:
- Market analysis & size
- Competitor breakdown
- Technical recommendations
- Cost estimates
5. Save output as `research-[YourAppName].txt`
**π‘ Pro Tip:** Use Gemini 2.5 Pro for better research (1,048,576-token context window).
### 2οΈβ£ Product Requirements (PRD) π
Define exactly what you're building β’ 15-20 min β’ Creates PRD-*.md
**What this does:** Transforms your idea into clear, actionable product specifications.
**How it works:**
1. Copy `part2-prd-mvp.md` into a new AI chat
2. Attach your research findings when prompted
3. Answer questions about:
- Core features (3-5 must-haves)
- Target users
- Success metrics
- UI/UX vision
4. AI creates professional PRD document
5. Save as `PRD-[YourAppName]-MVP.md`
### 3οΈβ£ Technical Design ποΈ
Plan the technical architecture β’ 15-20 min β’ Creates TechDesign-*.md
**What this does:** Decides the best tech stack and implementation approach for 2025.
**How it works:**
1. Copy `part3-tech-design-mvp.md` into a new AI chat
2. Attach your PRD (required) and research (optional)
3. Answer questions about:
- Platform (web/mobile/desktop)
- Complexity tolerance
- Budget constraints
- Timeline
4. AI recommends optimal stack from:
- No-code: Bolt.new, Lovable, Bubble
- Low-code: Next.js + Supabase
- Full-code: Your preferred framework
5. Save as `TechDesign-[YourAppName]-MVP.md`
### 4οΈβ£ Generate AI Agent Instructions π€
Create blueprints for your AI coding assistant β’ 5-10 min β’ Creates NOTES.md + agent configs
**What this does:** Converts all docs into step-by-step coding instructions for AI agents.
**How it works:**
1. Copy `part4-notes-for-agent.md` into a new AI chat
2. Attach PRD and Technical Design documents
3. AI generates:
- `NOTES.md` - Universal instructions
- Tool-specific configs (based on your choice):
- `CLAUDE.md` for Claude Code
- `GEMINI.md` for Gemini CLI
- `AGENTS.md` for Jules
- `.cursorrules` for Cursor
- `.windsurfrules` for Windsurf
4. Save all files in your project root
### 5οΈβ£ Build with AI Agent π»
Let AI build your MVP β’ 1-3 hrs β’ Creates working application
#### Setup by Tool Type
Terminal Agents (Claude Code, Gemini CLI)
```bash
# Claude Code
npm install -g @anthropic-ai/claude-code
cd your-project
claude init
# Add CLAUDE.md to project root
claude "Read CLAUDE.md and NOTES.md, then build the MVP"
# Gemini CLI
npm install -g @google-gemini/cli
gemini login
# Add GEMINI.md to project root
gemini "Read GEMINI.md and NOTES.md, then implement"
```
IDE Tools (Cursor, Windsurf)
1. Open your project folder in the IDE
2. Add configuration file:
- Cursor: `.cursorrules` or `.cursor/rules.mdc`
- Windsurf: `.windsurfrules`
3. Start with: *"Read NOTES.md and build the MVP step by step"*
No-Code Platforms (Bolt.new, Lovable)
1. Go to platform website
2. Paste your PRD content
3. Say: *"Build this MVP following the specifications"*
4. Deploy instantly with one click
#### Essential Prompts for Building
**Starting prompts by experience level:**
| Level | First Prompt |
|-------|--------------|
| **Beginner** | *"I'm new to coding. Read NOTES.md and guide me step-by-step to build this MVP. Explain what you're doing."* |
| **Intermediate** | *"Read NOTES.md and the docs folder. Build the core features first, test, then add polish."* |
| **Developer** | *"Review NOTES.md and architecture. Implement Phase 1 with proper patterns and test coverage."* |
**Follow-up prompts for all levels:**
- *"Show me the current progress vs requirements"*
- *"Test [feature] and fix any issues"*
- *"Add error handling and edge cases"*
- *"Generate README with setup instructions"*
- *"Prepare for deployment to [platform]"*
---
## π Final Project Structure
```
your-app/
βββ docs/
β βββ research-YourApp.txt
β βββ PRD-YourApp-MVP.md
β βββ TechDesign-YourApp-MVP.md
βββ NOTES.md # Universal AI instructions
βββ CLAUDE.md # Claude Code config (if using)
βββ GEMINI.md # Gemini CLI config (if using)
βββ AGENTS.md # Jules config (if using)
βββ .cursorrules # Cursor config (if using)
βββ .windsurfrules # Windsurf config (if using)
βββ README.md # Setup instructions (AI-generated)
βββ .env.example # Environment variables
βββ src/ # Your application code
```
---
## π What's New in 2025
Latest AI Models & Capabilities
### Model Updates
- **Claude Sonnet 4.5** β September 2025 release with 200K default / 1M beta context and top-tier architecture reasoning
- **Claude Opus (2025 snapshot)** β Handles massive monorepos and multi-hour planning sessions with improved tool use
- **Gemini 2.5 Pro** β 1,048,576-token input and 65,536-token output windows for deep research and synthesis
- **GPT-5** β Responses API with adjustable reasoning effort, faster tool orchestration, and lower latency
### New Tools
- **Claude Code** - Anthropic's project-aware terminal agent with session memory and automated test orchestration
- **Jules** - Google's asynchronous coding agent that can work independently across Vertex AI projects
- **Gemini CLI** - Open-source CLI with direct Gemini 2.5 Pro/Flash access, live context streaming, and MCP integrations
- **GitHub Copilot Agent Mode** - Workspace automation for scoped plans, PRs, and deployment checklists
- **OpenAI CLI (Responses API)** - Terminal workflow for GPT-5, o-series reasoning models, and tool-calling pipelines
### Platform Updates
- **Cursor** - Adds Codebase Q&A, long-term project Memory, and Auto Debug powered by Claude Sonnet 4.5
- **Windsurf** - Cascade AI 3.0 with worklog tracking and pair-mode for live coding
- **Cline** - MCP marketplace runners plus transparent execution summaries in VS Code
- **Bolt.new** - Instant Next.js/Supabase deployments with scheduled automations and $20M ARR milestone
- **Lovable** - AI fullstack builder shipping 25k apps daily after β¬14.3M growth round
---
## π‘ Pro Tips for 2025
Tool Selection Guide
| Persona | Best Tool Stack | Why it fits | What to watch | Setup time |
|---------|-----------------|-----------|--------------|------------|
| Complete beginner | Lovable β’ Bolt.new | Paste your idea, get a hosted app with a domain and database in minutes | Daily credit caps and hosted code mean you should harden it later | 10β30 min |
| Learning hobbyist | Copilot Agent (VS Code) β’ Cline | Copilot can edit files for you; Cline shows every diff and runs commands on request | Copilot Free has limits and Cline needs your API key setup | 20β45 min |
| Experienced developer | Cursor 2.0 β’ Windsurf | Cursor plans tasks and runs code in safe terminals; Windsurf handles bigger refactors with its agents | Usage-based credits and a new IDE to learn | 30β60 min |
| Budget-limited builder | Cline β’ Gemini CLI | Both are free to install, work locally, and can call Gemini 2.5 without paid tiers | Less hand-holding than no-code toolsβexpect to prompt more | 15β40 min |
| Need-it-today founder | Lovable Agent Mode β’ Bolt.new (Claude) | Fastest path to a working MVP with dashboards, auth, and analytics baked in | Keep an eye on credit burn and schedule a security/UX review | 15β60 min |
| Mobile-first product team | v0.dev + v0 Mobile β’ Flutter + Gemini | v0 sketches mobile experiences; Flutter + Gemini in Android Studio builds native quality | v0 mobile features are new and Android Studio tooling is still evolving | 45β120 min |
| Complex logic engineer | Claude Code (web/VS Code) β’ Windsurf | Claude Sonnet 4.5 keeps huge context in its βmemory,β Windsurfβs agents plan multi-file changes | Claude web is still preview and company data rules may restrict it | 30β90 min |
| Security/compliance lead | Cline (client-side) β’ Copilot Enterprise | All code stays local with Cline, while Copilot Enterprise adds SSO and audit logs | You still need team policies and Copilot request caps may apply | 60β120 min |
| Offline/privacy-focused dev | Gemini CLI (local agent) β’ Cline + local models | Works from your machine, and you can swap in local models (Ollama/DeepSeek) when needed | Fully offline mode depends on your hardware and chosen model | 30β60 min |
| Open-source maintainer | Cline β’ Aider | Both tools show diffs, commit for you, and play nicely with Git workflows | Terminal-first experience can feel advanced if Git basics are new | 20β45 min |
**Quick Picks (Plain English)**
- Need an MVP tonight? Use Lovable or Bolt.new, then plan a follow-up pass for polish and security.
- Working inside an editor? Cursor or Copilot Agent can code for youβopen Cline when you want to double-check every change.
- Huge legacy repo? Bring in Claude Code or Windsurf so the agent βremembersβ more of your code at once.
- No budget? Stick with Cline + Gemini CLI and the generous free request limits.
- Mobile-first? Preview flows in v0 Mobile or build native screens with Flutter + Geminiβs suggestions.
**Pricing Guardrails (Nov 2025)**
- Free forever: Gemini Code Assist (6K code reqs/day), Copilot Free (2K completions + 50 chats), Windsurf Free (25 prompts).
- Best under $20: Windsurf Pro $15/mo (500 credits), Copilot Pro $10/mo (Agent Mode), Cursor Pro $20/mo (credit pool + parallel agents).
- Power tiers: Copilot Pro+ $39/mo (1.5K premium calls), Cursor Ultra $200/mo (20Γ credits), Bolt Pro 200 $200/mo (120M tokens).
**MCP Watch**
- Model Context Protocol (MCP) is a shared standard that lets your AI talk to extra tools (think CI pipelines, databases, ticket queues).
- Cline, Cursor, Claude Code, and Gemini CLI all plug into MCP, but treat those connections like production credentials.
**Caution Flags**
- Claude consumer accounts now offer opt-in data sharing; stick to org workspaces or API keys for sensitive repos.
- Replit Agent 3 can run 200-minute autonomous loopsβenable guardrails so retries cannot nuke production data.
**When Not to Use These Tools**
- Native mobile or hardware builds: prefer Flutter + traditional toolchains with AI pairing.
- Regulated workloads needing SOC2/FedRAMP/HIPAA: lean on Copilot Enterprise, Cline with strict policies, or self-hosted stacks.
- Safety-critical or real-time systems: require deterministic, human-led engineering.
- Fundamentals practice: hand-code portions to avoid skipping core concepts.
Common Pitfalls & Solutions
| β Pitfall | β
Solution |
|-----------|------------|
| Skipping discovery work | Run the Part I research prompt first so the PRD and tech design arenβt guesses |
| Letting agents ship code alone | Ask agents to show their plan, review the diff, and run tests before anything merges |
| Sticker-shock bills | Check credit dashboards weekly and keep a backup stack (Cline + Gemini) that stays free |
| Publishing auto-generated UIs without checks | Test accessibility, security, and performance before launch day |
| Building sensitive apps with personal accounts | Use business plans or API workspaces, switch off data sharing, and keep secrets in vaults |
| Forcing one tool to do everything | Mix and match (IDE + terminal + builder) so each tool covers what it does best |
---
## π¨ Troubleshooting
Quick Fixes for Common Issues
| Problem | Solution |
|---------|----------|
| **"AI ignores my documents"** | Start with: *"First read NOTES.md, PRD, and TechDesign. Summarize key requirements before coding."* |
| **"Code doesn't match PRD"** | Say: *"Re-read the PRD section on [feature], list acceptance criteria, then refactor accordingly."* |
| **"AI is overcomplicating"** | Add to config: *"Prioritize MVP scope. Offer the simplest working implementation before optimizations."* |
| **"Lost track of progress"** | Ask: *"Update the NOTES.md progress log and map remaining tasks to implementation phases."* |
| **"Deployment failing"** | Request: *"Walk through deployment checklist, verify env vars, then run the platform-specific health command."* |
---
## π€ Contributing
PRs & issues welcome! Help us improve:
- π Report issues with prompts
- π‘ Share your success stories
- π§ Add new tool configurations
- π Submit example MVPs built with this workflow
---
## π License
Released under the [MIT License](LICENSE).
---
**The best time to build your idea was yesterday.**
**The second best time is now.** π
*Last updated: November 2025 | Created by the vibe-coding community*