很多初学者都用 AI 写代码，但项目报错时不会自己调，又是反复问 AI，无法快速定位问题。

ownyourcode，一个让 AI 当导师不当码农的工具，只负责引导、提问、审查，代码还是我们自己敲。

每个开发任务完成前，需过六道质量关卡，逼着我们真的搞懂代码，而不是复制粘贴了事。

GitHub：http://github.com/DanielPodolsky/ownyourcode…

按学习阶段分了初级、转行、面试准备几档，教学方式跟着调整。

想借 AI 提速、又不想把脑子全交出去的开发者，可以拿试下它。

DanielPodolsky/ownyourcode

Source: https://github.com/DanielPodolsky/ownyourcode

AI-Mentored Development
AI guides, you build. You own the result.

_{v2.7.0 · MIT License}

The Problem

AI coding tools optimize for shipping. The risk: code that works but isn’t truly yours. Hard to debug without context. Hard to extend without asking again. Hard to defend when it matters.

OwnYourCode flips this. AI becomes your mentor, not your coder. It guides, questions, and reviews — but you write the code.

The result: Code you understand. Code you can extend. Code that’s actually yours.

Who This Is For

Profile selection is part of /own:init. Choose your profile and OwnYourCode adapts its teaching style.

Profiles adapt HOW we teach. The core (6 Gates, code reviews, quality) stays the same. Change anytime with /own:profile.

Quick Start

macOS / Linux

curl -sSL https://raw.githubusercontent.com/DanielPodolsky/ownyourcode/main/scripts/base-install.sh | bash
cd your-project && ~/ownyourcode/scripts/project-install.sh

Windows (PowerShell)

irm https://raw.githubusercontent.com/DanielPodolsky/ownyourcode/main/scripts/base-install.ps1 | iex
cd your-project
irm https://raw.githubusercontent.com/DanielPodolsky/ownyourcode/main/scripts/project-install.ps1 | iex

Initialize

/own:init

How It Works

The 4 Protocols

The 6 Gates

Before completing any task, your code passes through 6 quality checkpoints:

Gate 1 blocks completion. Can’t explain your code? Don’t ship it.

The Flywheel

Learnings compound across projects. Patterns that worked. Mistakes you won’t repeat. Career value extracted from every task.

Commands

Core Workflow

Junior profile note: implementation for juniors happens inside /own:feature (Phase 6) as a predict-before-reveal loop — you commit a prediction before the AI reveals code, then get graded on it. /own:guide remains available for ad-hoc help but is no longer the primary path to writing code.

Utilities

MCP Setup (Recommended)

MCPs provide real-time documentation and production code examples.

# Context7 — Official documentation lookup
claude mcp add context7 --transport http https://mcp.context7.com/mcp

# Octocode — GitHub code search
https://octocode.ai/#installation

Without MCPs, OwnYourCode operates at reduced capability — guidance relies on AI’s training data rather than verified, up-to-date documentation.

Philosophy

“Won’t this slow me down?”

Yes. That’s the point.

Building with someone else’s code means you can’t build the next thing alone. Building yourself takes longer — but now you can build anything.

Full Philosophy →

The Dashboard

Your project’s plan isn’t scattered across a dozen Markdown files — it’s a single dashboard you open in your browser. Mission, stack, and roadmap up top; every phase opens Spec / Design / Tasks tabs (acceptance criteria, a real architecture diagram, a kanban board with a live progress ring). The /own:* commands keep it in sync as you work.

How it stays in sync: all state lives in one data file (ownyourcode/dashboard/dashboard-data.js); dashboard.html is a stable view that renders it. No server — you double-click the file. Want a different look? /own:theme re-skins it from a design brief.

_{MIT License}