Baku Skills: An AI Programming Assistant Skill Library Built for Chinese Developers

Baku Skills is an open-source, Chinese-first collection of Agent skills developed by Basic-XYZ on GitHub, designed for AI programming assistants like Codex and Claude Code. It provides best practices for coding standards, design systems, and development workflows. This project aims to address the problem of understanding biases in Chinese development scenarios caused by AI programming assistants' English-dominated training data. It supports scenarios such as standardized team collaboration and automated code reviews, and is open-source with encouragement for community contributions.

With the rapid development of AI programming assistants (such as GitHub Copilot, Claude Code, Codex, etc.), more and more developers rely on these tools. However, most AI assistants' training data is dominated by English, leading to understanding biases or cultural differences when handling Chinese development scenarios. The Baku Skills project was born as a Chinese-first collection of Agent skills, providing carefully designed prompt templates, coding standards, and workflow guidelines to help AI assistants better adapt to the Chinese development environment.

Baku Skills adopts a modular design and includes multiple skill packages:

1. baku-coding-discipline (Coding Standards)：Defines complete coding discipline and optimizes Chinese developers' habits (e.g., Chinese semantic mapping for variable naming, Chinese comment standards, etc.);
2. baku-design-system (Design System)：Provides reusable design standards (color systems, component design, etc.);
3. baku-illustrations (Illustration Standards)：Defines usage standards for illustrations and icons;
4. codex-history-recovery (History Recovery)：A practical tool module to solve the problem of AI conversation history loss.

The project uses skills.sh.json as the skill definition file (JSON format describing metadata, dependencies, etc.), which is convenient for AI assistants to automatically discover and load. It provides a convenient installation script (tools in the bin directory) and supports dependency management for the Node.js ecosystem via package.json. It also provides English documentation (README.en.md) to balance Chinese-first priority and the needs of international developers.

Baku Skills is designed with compatibility for mainstream AI assistants in mind:

• Codex (OpenAI)：Optimizes prompt engineering to ensure accurate understanding of skill definitions;
• Claude Code (Anthropic)：Modular design fits its Agent architecture and can be loaded via extension mechanisms;
• General Workflow：Supports adoption by multiple tools with an open design philosophy.

The practical value of Baku Skills includes:

• Standardized Team Collaboration：Serves as a benchmark for coding standards, helping new members quickly learn team standards;
• Automated Code Review：AI assistants automatically check standards, reducing manual workload;
• New Member Training：Provides a structured learning path;
• Cross-Project Consistency：Centralized skill definition updates benefit all projects.

Baku Skills is open-source under the MIT license and encourages community contributions. Contribution methods include submitting new skill modules, improving existing definitions, perfecting documentation examples, reporting issues and suggestions, etc. The project has ecological integration potential and can be adopted by individuals or enterprises, or integrated into the official ecosystems of AI assistant vendors.

Baku Skills represents the direction of AI-assisted programming from general capabilities to domain-specific specialization. It can fill model knowledge gaps, solidify best practices, improve generation quality, and promote knowledge sharing. As a pioneering attempt in the Chinese developer community, it has reference value for teams introducing AI-assisted programming, and similar skill library ecosystems will flourish in the future.