Maker Flow

A personal MVP incubation pipeline

Heavy infrastructure, light logic. Kill repeated setup so the path from idea → public validation has minimal friction.

You provide requirements and two approvals. AI agents execute against a skill library and template set.

English · 简体中文

Getting started · Live example · Template catalog · Skills catalog · For agents · Docs index

---

Why it exists

For indie builders, friction is often not business code — it is rebuilding the environment every time:

Without Maker Flow	With Maker Flow
Pick a stack, write Docker, wire Nginx for every idea	Templates ready to run
AI codes first; wrong direction wastes the day	Two gates: PRO → MVP
Prompts and deploy steps live in your head	Skills encode SOPs; agents follow them
Many ideas, repeated plumbing	Focus on validation; public URL in ~10 minutes

This is not one product. It is a forkable, reusable MVP factory.
Recommended: keep this repo as the public tool; build each MVP in a separate private product repo → consumer guide.

---

Six-step pipeline

flowchart LR
    A["① You: one-line requirement"] --> B["② AI: draft PRO"]
    B --> C{"③ Approve PRO"}
    C -->|pass| D["④ AI: match templates · assemble MVP"]
    C -->|revise| B
    D --> E{"⑤ Approve MVP"}
    E -->|pass| F["⑥ Deploy"]
    E -->|iterate| D
    F --> G["🌐 Public traffic"]

Loading

Step	You	AI agent
1	Provide requirement	—
2	—	Draft PRO (no code)
3	Approve PRO	Wait
4	—	Match templates → assemble in product repo
5	Local acceptance	Fix against PRO
6	Confirm where to publish	Ask in chat → follow skills/publish-<target>.md

Two gates are the core design: align on what first, then how.

---

What’s in the repo

        ┌─────────────┐
        │  You + AI   │
        └──────┬──────┘
               │
    ┌──────────┼──────────┐
    ▼          ▼          ▼
 skills/   templates/   release/
  (HOW)      (WHAT)      (SHIP)

Module	Path	One line
Skills	skills/ · catalog	How agents draft PRO, match templates, publish
Templates	templates/ · catalog	apps + images + patterns
Release	release/	Step-6 scripts (VPS deploy, Nginx gateway, Cloudflare DNS CLI)
Optional LLM	docs/optional-llm.md	Rare: self-hosted OpenAI-compatible APIs

---

60-second start

curl -fsSL https://raw.githubusercontent.com/LJTian/maker-flow/main/scripts/install.sh | bash
maker-flow new my-first-mvp
cd ~/projects/my-first-mvp

Open the product repo in an AI agent IDE, @AGENTS.md, start at step ①.

Contributors: install from a git clone

git clone https://github.com/LJTian/maker-flow.git && cd maker-flow
./scripts/install.sh
maker-flow new my-first-mvp

A — AI Agent (recommended)

1. Open the product repo (~/projects/<name>/) in an AI agent IDE

2. Start a chat:

   Follow @AGENTS.md. I want to build [your idea], starting at step ①.

3. Approve PRO and MVP at steps ③ and ⑤

Full guide → docs/getting-started.md · 中文

Live example: static intro with web-vite → GitHub Pages — walkthrough · demo

---

Who it’s for

• People with many ideas who want fast public feedback
• Anyone using Cursor / Claude agents who is tired of prompting from scratch
• Builders who want a forkable private pipeline, not another todo demo
• Anyone who buys heavy infra, light logic: write plumbing once, run N ideas

---

After you Star / Fork

Action	Suggestion
Star	Track skill / template updates
Fork / clone	Use as the shared factory (public)
Each new idea	New private product repo + consumer guide (maker-flow new <name>)
Pin agent behavior	Factory: AGENTS.md · Product repo: AGENTS.consumer.example.md

---

Docs map

Humans	AI agents
Getting started	AGENTS.md
Consumer project · Overview	workflow.md
Template catalog · Skills catalog	agent-bootstrap.md
Docs index · i18n	AGENTS.consumer.example.md (product repos)

---

Acknowledgments

Maker Flow stands on these projects — thanks to maintainers and communities:

Use	Project	URL
Go web (go-api)	Gin	https://github.com/gin-gonic/gin
Web UI (web-vite)	Vite · React · Tailwind CSS	https://vite.dev/ · https://react.dev/ · https://tailwindcss.com/
CLI (go-cli)	Cobra	https://github.com/spf13/cobra
singleflight (pattern)	golang.org/x/sync	https://pkg.go.dev/golang.org/x/sync/singleflight
Build fragment (Go)	golang (official)	https://hub.docker.com/_/golang
Runtime fragment (Alpine)	Alpine Linux	https://alpinelinux.org/ · https://hub.docker.com/_/alpine
Containers	Docker	https://www.docker.com/
Reverse proxy	Nginx	https://nginx.org/
DNS / edge SSL	Cloudflare	https://www.cloudflare.com/
Optional local LLM	Ollama	https://ollama.com/ · https://github.com/ollama/ollama
Agent protocol reference	Model Context Protocol	https://modelcontextprotocol.io/

Missing a credit? Open an Issue / PR.

---

If this pipeline helps you, a Star is the best nod to “heavy infrastructure.”

Get started · Fork your factory

License

MIT — use freely and adapt as needed.