A structured prompt that turns any AI into your technical co-founder.
Not just code generation — a 5-stage workflow from idea to shipped product.
中文版 · Get the Prompt · Live Demo
- What is this?
- The 5-Stage Workflow
- Quick Start
- What's Inside
- Example Conversation
- Cases
- Who is this for?
- FAQ
- Contributing
- License
Most AI prompts give you a personality. This one gives you a process.
| Approach | What you get |
|---|---|
| "Build me an app" | Code (maybe broken, maybe not what you need) |
| AI character prompts | A fun conversation, no real output |
| Vibe to Ship | A structured workflow with quality gates at every stage |
You don't just get an app. You learn how to think like a product owner.
① Explore → ② Plan → ③ Build → ④ Polish → ⑤ Ship
| Stage | What happens | Who decides |
|---|---|---|
| Explore | AI asks probing questions to understand the real problem | You answer, AI challenges assumptions |
| Plan | Define features, tech stack, complexity, resources | You approve the plan |
| Build | Vertical slices, each tested before moving on | You review each slice, choose at forks |
| Polish | Design consistency, edge cases, performance, accessibility | You approve the final look |
| Ship | Deploy, document, maintain | You own the product |
Each stage has quality gates — you can't move forward until the previous stage is solid.
- Copy the prompt from
prompts/cofounder.md - Paste into ChatGPT / Claude / Cursor / any AI
- Tell it what you want to build
- Follow the 5-stage workflow
# Copy the skill to your Trae skills directory
cp -r skills/tech-cofounder ~/.trae/skills/Then in Trae, just say "I want to build a tool" and the AI will automatically enter co-founder mode.
# Copy the skill to your Claude Code skills directory
cp -r skills/tech-cofounder ~/.claude/skills/
# Copy rules to your Claude Code rules directory
cp -r rules/* ~/.claude/rules/
# Copy agents to your Claude Code agents directory
cp -r agents/* ~/.claude/agents/
# Copy commands to your Claude Code commands directory
cp -r commands/* ~/.claude/commands/Vibe to Ship is now a complete system — not just a prompt, but a modular framework with rules, agents, commands, and skills.
vibe-to-ship/
├── prompts/
│ └── cofounder.md # The prompt — copy & paste into any AI
├── skills/
│ ├── tech-cofounder/
│ │ └── SKILL.md # Trae/Claude Code skill (auto-activates)
│ ├── brainstorming/
│ │ └── SKILL.md # Design refinement with HARD-GATE
│ ├── writing-plans/
│ │ └── SKILL.md # Detailed implementation plans
│ ├── subagent-driven-development/
│ │ └── SKILL.md # Execute plans with fresh subagents
│ ├── test-driven-development/
│ │ └── SKILL.md # RED-GREEN-REFACTOR cycle
│ ├── verification-before-completion/
│ │ └── SKILL.md # Ensure work is actually done
│ ├── requesting-code-review/
│ │ └── SKILL.md # Structured review process
│ ├── receiving-code-review/
│ │ └── SKILL.md # Respond to feedback constructively
│ ├── model-selection/
│ │ └── SKILL.md # Choose right model for each task
│ └── context-window-management/
│ └── SKILL.md # Optimize context usage
├── rules/ # Always-follow guidelines
│ ├── security.md # Security checklist (no secrets, input validation)
│ ├── coding-style.md # Immutability, file org, naming, error handling
│ ├── testing.md # TDD workflow, 80% coverage, edge cases
│ └── quality.md # Stage gates, visual polish, accessibility
├── agents/ # Specialized sub-agents for delegation
│ ├── planner.md # Implementation planning before coding
│ ├── code-reviewer.md # Quality, security, and maintainability review
│ ├── security-reviewer.md # OWASP Top 10 and vulnerability detection
│ └── polish-agent.md # Visual polish, edge cases, responsive design
├── commands/ # Quick shortcuts for common tasks
│ ├── plan.md # /plan — Create implementation plan
│ ├── review.md # /review — Code quality review
│ ├── security-check.md # /security-check — Security audit
│ ├── test.md # /test — Run tests and check coverage
│ ├── polish.md # /polish — Visual polish and edge cases
│ └── ship.md # /ship — Pre-deployment checks
├── examples/
│ └── habit-streak/
│ └── index.html # Live demo — open in browser
├── CONTRIBUTING.md
├── CODE_OF_CONDUCT.md
├── LICENSE
└── README.md
Modular guidelines that ensure every product is secure, well-coded, and tested:
| Rule | What it covers |
|---|---|
| Security | No hardcoded secrets, input validation, OWASP basics, emergency protocol |
| Coding Style | Immutability, small files, naming conventions, error handling |
| Testing | TDD workflow, 80% coverage, edge case testing, test structure |
| Quality | Stage gates, visual consistency, responsive design, accessibility |
Specialized sub-agents you can delegate tasks to:
| Agent | When to use |
|---|---|
| Planner | Before writing code — creates detailed implementation plans |
| Code Reviewer | After writing code — reviews quality, security, patterns |
| Security Reviewer | Auth/API/payment code — OWASP audit, vulnerability detection |
| Polish Agent | Stage 4 — visual consistency, edge cases, responsive, a11y |
Slash commands for quick access to common workflows:
| Command | Purpose |
|---|---|
/plan |
Create implementation plan before coding |
/review |
Code quality and security review |
/security-check |
Focused security audit |
/test |
Run tests and check coverage |
/polish |
Visual polish and edge case handling |
/ship |
Pre-deployment checks and deployment |
Modular skills that enhance each stage of the workflow:
| Category | Skill | Purpose |
|---|---|---|
| Design | Brainstorming | Design refinement through Socratic dialogue (HARD-GATE) |
| Design | Writing Plans | Create detailed implementation plans |
| Design | Subagent Development | Execute plans with fresh subagents per task |
| Development | Test-Driven Development | RED-GREEN-REFACTOR cycle |
| Development | Verification | Ensure work is actually done with evidence |
| Review | Requesting Code Review | Structured review process |
| Review | Receiving Code Review | Respond to feedback constructively |
| Optimization | Model Selection | Choose right model for each task |
| Optimization | Context Window Management | Optimize context usage |
You: I want a tool to track my habits
AI (as co-founder): Good idea. Before I build anything, let me understand:
1. How do you currently track habits? What's broken about that?
2. Is this just for you, or do you want to share it?
3. What's the smallest version that would still be useful?
4. What's your timeline?
You: I use a spreadsheet. It's ugly and I forget to update it.
Just for me. I just need a streak counter and a daily reminder.
This weekend.
AI: Perfect. Here's my plan for v1:
- Single-file HTML (no server, no install)
- One habit, one streak, one reminder
- Neumorphic design, mobile-first
- Complexity: simple (2-3 hours)
Shall I proceed?
A minimal habit tracker built with the Vibe to Ship workflow — from idea to polished demo in a single session.
| What it does | Track one habit, visualize streaks with a GitHub-style heatmap, AI-powered check-in suggestions |
| Tech | Single-file HTML · localStorage · Zero dependencies · Neumorphic design |
| Built in | ~2 hours, 1 session |
Features: Confetti celebrations at milestones · Dark mode · Export JSON/CSV · Mobile-first responsive · Data stays on device
Want to contribute your case? See CONTRIBUTING.md.
- Solo builders who want to go from idea to product fast
- Non-technical founders who want to understand what they're asking AI to build
- Developers who want a structured workflow instead of ad-hoc prompting
- Product managers who want to prototype ideas before handing off to engineering
Q: Is this just a prompt? A: It's a structured workflow embedded in a prompt. The prompt guides the AI through 5 stages with decision points and quality gates.
Q: Does it work with GPT-4 / Claude 3.5 / etc? A: Yes. It works with any AI that can follow structured instructions. Tested with ChatGPT, Claude, Cursor, and Trae.
Q: Do I need to know how to code? A: No. The AI handles all code. You make product decisions. But you'll learn some code along the way.
Q: What if my idea is too big? A: The Explore stage will help you find the smallest valuable slice. The AI will challenge scope creep.
Q: Can I use this for non-software projects? A: The workflow (Explore → Plan → Build → Polish → Ship) works for any creative project. The code-specific parts can be skipped.
We welcome contributions! See CONTRIBUTING.md for guidelines.
This project is licensed under the MIT License — see LICENSE for details.
If this project helped you build something, consider giving it a ⭐
Built something with this prompt? Open an issue and share it!
Suggested GitHub Topics: vibe-to-ship · prompt-engineering · vibe-coding · ai-workflow · system-prompt · product-management