start-vibe-coding

Drop INIT.md into a fresh repo, tell your agent what you want to build, and let it start vibe coding with real project context from the first pass.

English | 简体中文

---

Why This Exists

Starting a project with an agent usually means repeating the same setup loop:

• write the first README
• explain the project structure
• define coding rules
• add environment setup notes
• restate the same context every time the repo is new

start-vibe-coding turns that repetition into a reusable protocol. One file gives the agent enough durable context to create a clean project surface instead of waiting for more prompt-by-prompt instruction.

---

What This Is

start-vibe-coding is a lightweight repository bootstrap protocol centered around INIT.md.

It helps an agent initialize a new repository with:

• human-facing docs
• agent-facing operating rules
• sensible repo defaults
• a minimal working structure that matches the project you described

---

How To Use It

1. Drop INIT.md into your new repository

Copy the protocol into the root of the project you want to bootstrap.

2. Tell the agent what you want to build

Describe the product, tool, app, script, or library in plain language.

3. Ask the agent to initialize the repo from the protocol

Example prompt:

I want to build a personal expense tracker web app.
Initialize this project following INIT.md.

From there, the agent should be able to:

• create the required files
• populate them with project-specific content
• set up the default local environment baseline
• establish a minimal working structure
• keep the human-facing and agent-facing docs consistent

---

What INIT.md Creates

The protocol defines a compact starter set:

• README.md for the project overview and usage
• PROJECT_LOG.md for append-only development history
• SPEC.md for the project map
• AGENTS.md for durable coding and workflow rules
• CLAUDE.md as a lightweight agent entry point
• .gitignore for local and generated files that should stay untracked
• requirements.txt as the default dependency manifest
• LICENSE for explicit licensing from day one

By default, the protocol also expects a repository-root .venv so Python tooling, dependency installation, and agent instructions all point at the same local environment.

---

Why It Works Well

This protocol creates a clean split between documents for people and documents for agents.

Audience	Files	Job
Humans	README.md, PROJECT_LOG.md	Explain the project, setup, usage, and progress
Agents	SPEC.md, AGENTS.md, CLAUDE.md	Define structure, rules, and execution expectations

That separation makes collaboration smoother:

• humans get a familiar project surface
• agents get explicit operating instructions
• both stay aligned as the repo evolves

---

Why The Agent Files Matter

Most projects already have a README. Fewer have durable instructions for coding agents.

start-vibe-coding treats agent-facing files as first-class project infrastructure:

• AGENTS.md holds the durable working rules
• CLAUDE.md stays minimal and points the agent at those rules
• SPEC.md keeps the repo map easy to reload

This gives the agent a stable operating context instead of relying on repeated chat prompts.

If needed, the same protocol can support multilingual docs such as README.zh.md.

---

License

MIT License