Vibe Coding CN 🧩 Skill Open source

What it does

Vibe Coding CN is a Chinese-language knowledge base and skill pack for AI-assisted development. Rather than a single tool, it packages a methodology the project calls 道法术器 (Tao-Fa-Shu-Qi: philosophy, method, process, tools) for turning ideas into maintainable products with an AI coding agent, built around Quality Gates — tests, CI/CD, type checking, schemas and checklists that constrain what the agent is allowed to ship.

What is inside

• docs/: a structured knowledge base covering vibe-coding concepts, workflows and philosophy, including six core propositions about model capability (fixed targets with variable strategies, isolated review to prevent self-confirmation bias, capability orchestration over novel generation, and more)
• prompts/: a prompt library (linked to a companion spreadsheet) covering requirement clarification, planning, execution and meta-prompts for generating further prompts
• skills/: reusable agent skills, including auto-skill (generates, refactors and validates Skills themselves), auto-tmux (scriptable terminal multiplexing so multiple AI sessions become observable and schedulable), and official Claude skill integrations
• tools/: Codex CLI configuration and helper scripts
• research/: write-ups tracking emerging agentic-engineering patterns, including recursive self-optimizing prompt systems

Works with

The repo is written around Codex CLI as its "default local execution entry point", but the docs and prompt templates are designed to be pasted into any web or CLI assistant (ChatGPT, Claude, Gemini) to generate an OS-specific setup, so the material isn't locked to one vendor.

How to install or connect

There's no package to install; clone the repository and read the docs, or copy individual skills into your agent's skills folder:

git clone https://github.com/tradecatlabs/vibe-coding-cn

Maintenance and safety

Active and fast-growing (16,900+ stars, weekly commits, MIT licensed). As with any community prompt/skill collection, read a skill before dropping it into an agent that can run shell commands.

Who should use it

Developers who are new to structured AI-assisted programming, and Chinese-speaking teams specifically who want a systematic, quality-gated workflow — with explicit human/AI responsibility boundaries and a real philosophy behind it — rather than ad hoc prompting.