Trae Agent Rules (Learning Edition)

中文 | English

这是一个关于 Trae IDE 智能体规则配置的学习项目，旨在通过标准化、结构化的规则定义，提升 AI 辅助编程的效率与准确性。

⚠️ 声明 (Disclaimer)

本项目是我 (@Cooper-X-Oak) 的学习之作。内容深受开源社区 Agent Rules 相关项目的启发，在此基础上进行了重构、汉化和定制，以适应 Trae IDE 的工作流。

✨ 项目亮点 (Highlights)

本项目对 Agent Rules 进行了深度的标准化和结构化改造，主要特点如下：

1. 自我指向的递归进化 (Self-Referential Evolution)

   • 核心哲学：本项目构建了一个“自我指向”的文档系统，即利用规则系统本身来规范和完善规则的制定。
   • 极致精简：采用 Key:Value 键值对格式，摒弃冗长描述，实现 Token 的最大化利用。
   • 强指令约束：使用 MUST, NEVER, REQUIRED 等关键词，降低理解歧义。

2. 智能生效模式 (Smart Triggering)

   • alwaysApply: true: 全局生效（如中文交互规范）。
   • globs: "*.md": 基于文件路径自动匹配生效。
   • description: "...": 基于语义场景智能触发。

3. 标准化分类命名 (Standardized Categories)

   • 所有规则文件采用 Category-功能名称.md 格式，分类清晰（Global/Core/Docs/Tech/Tools）。

📂 目录结构

所有规则文件均位于 TRAERULES/ 目录下，采用扁平化管理。新版架构引入了元规则（Meta-Rules）来增强系统的自我管理能力：

• _Meta-规则架构标准.md: [核心元规则] 定义了所有规则必须遵循的宪法级标准（替代了旧版的 RULEOFRULES.md）。
• _Meta-规则索引专家.md: [索引与引导] 全局规则索引，负责根据用户意图引导加载特定的领域规则。
• Core (核心类)
  • Core-中文交互规范.md: 全局语言约束，确保中文回复。
  • Core-Bug修复专家.md: 标准化 Bug 修复流程。
  • Core-技术实现架构师.md: 代码实现与架构设计。
  • ... (全视角审查、根因分析等)
• Docs (文档类)
  • Docs-技术文档工程师.md: 编写技术文档。
  • Docs-版本日志管理员.md: 维护 CHANGELOG.md。
  • ... (文档维护、上下文引导等)
• Tech (技术栈类)
  • Tech-现代Swift专家.md: Swift 开发规范。
  • Tech-浏览器自动化工程师.md: 浏览器自动化开发。
• Tools (工具类)
  • Tools-Github项目管家.md: Github 项目管理。
  • Tools-GitCommit专家.md: 规范化 Git 提交。
  • Tools-脚本工具规范.md: [NEW] 规范临时脚本的创建与管理。
  • ... (MCP配置、系统可视化等)

🚀 如何使用 (Usage)

本仓库只包含整理好的规则文件，不包含 Trae 的项目配置或原始素材。

1. 获取规则

git clone https://github.com/Cooper-X-Oak/trae-agent-rules.git

2. 集成到项目

1. 在你的 Trae 项目根目录下，创建 .trae/rules/ 文件夹。
2. 将 TRAERULES 文件夹中的所有 .md 文件复制到 .trae/rules/ 目录下。
3. 重要：确保 _Meta-规则架构标准.md、_Meta-规则索引专家.md 和 Core-中文交互规范.md 被包含在内，以保证基础体验。

3. 验证生效

在 Trae 中打开任意文件或进行对话，AI 将根据当前的上下文（文件类型或你的问题）自动加载相应的规则。例如，询问“项目结构”时应触发 Docs-项目上下文引导员。

🤝 贡献 (Contributing)

欢迎提交 Issue 或 Pull Request 来改进规则库！

1. Fork 本仓库。
2. 创建特性分支 (git checkout -b feature/AmazingRule).
3. 按照 _Meta-规则架构标准.md 的标准编写新规则。
4. 提交更改 (git commit -m 'feat: Add AmazingRule').
5. 推送到分支 (git push origin feature/AmazingRule).
6. 提交 Pull Request。

📄 许可证 (License)

MIT License

---

Created with ❤️ by Cooper-X-Oak using Trae IDE

---

Trae Agent Rules (Learning Edition)

中文 | English

This is a learning project about Trae IDE agent rule configuration, aiming to improve the efficiency and accuracy of AI-assisted programming through standardized and structured rule definitions.

⚠️ Disclaimer

This project is a learning work by me (@Cooper-X-Oak). The content is deeply inspired by Agent Rules related projects in the open-source community, and has been refactored, localized, and customized to fit the Trae IDE workflow.

✨ Highlights

This project has undergone deep standardization and structural transformation of Agent Rules. Key features include:

1. Self-Referential Evolution

   • Core Philosophy: This project builds a "self-referential" documentation system, meaning it uses the rule system itself to regulate and improve rule creation.
   • Extreme Minimalism: Adopts Key:Value format, discarding verbose descriptions to maximize Token utilization.
   • Strong Constraint Directives: Uses keywords like MUST, NEVER, REQUIRED to reduce ambiguity.

2. Smart Triggering

   • alwaysApply: true: Globally effective (e.g., Language Constraint).
   • globs: "*.md": Automatically triggered based on file paths.
   • description: "...": Intelligently triggered based on semantic context.

3. Standardized Categories

   • All rule files follow the Category-FeatureName.md format with clear classification (Global/Core/Docs/Tech/Tools).

📂 Directory Structure

All rule files are located in the TRAERULES_EN/ directory (for English version), managed in a flat structure. The new architecture introduces Meta-Rules to enhance the system's self-management capabilities:

• _Meta-RulesArchStandard.md: [Core Meta-Rule] Defines the constitutional standards that all rules must follow (Replaces the old RULEOFRULES.md).
• _Meta-RulesIndexExpert.md: [Index & Guide] Global rule index, responsible for guiding the loading of specific domain rules based on user intent.
• Core
  • Core-LanguageConstraint.md: Global language constraint (Enforces English).
  • Core-BugFixExpert.md: Standardized Bug fix process.
  • Core-TechImplementationArchitect.md: Code implementation and architecture design.
  • ... (Comprehensive Review, Root Cause Analysis, etc.)
• Docs
  • Docs-TechWriter.md: Writes technical documentation.
  • Docs-ChangelogManager.md: Maintains CHANGELOG.md.
  • ... (Doc Maintainer, Context Guide, etc.)
• Tech
  • Tech-ModernSwiftExpert.md: Swift development standards.
  • Tech-BrowserAutomationEng.md: Browser automation development.
• Tools
  • Tools-GithubProjectManager.md: Github project management.
  • Tools-GitCommitExpert.md: Standardized Git commits.
  • Tools-ScriptToolSpec.md: [NEW] Standardize temporary script management.
  • ... (MCP Config, System Visualization, etc.)

🚀 Usage

This repository only contains the organized rule files and does not include Trae's project configuration or raw materials.

1. Get Rules

git clone https://github.com/Cooper-X-Oak/trae-agent-rules.git

2. Integrate into Project

1. In your Trae project root directory, create a .trae/rules/ folder.
2. Copy all .md files from the TRAERULES_EN folder to the .trae/rules/ directory.
3. Important: Ensure _Meta-RulesArchStandard.md, _Meta-RulesIndexExpert.md and Core-LanguageConstraint.md are included to guarantee the basic experience.

3. Verify

Open any file or start a conversation in Trae. The AI will automatically load the corresponding rules based on the current context (file type or your question). For example, asking about "project structure" should trigger Docs-ProjectContextGuide.

🤝 Contributing

Issues and Pull Requests are welcome to improve the rule library!

1. Fork this repository.
2. Create a feature branch (git checkout -b feature/AmazingRule).
3. Write a new rule following the standard in RULEOFRULES.md.
4. Commit changes (git commit -m 'feat: Add AmazingRule').
5. Push to branch (git push origin feature/AmazingRule).
6. Submit a Pull Request.

📄 License

MIT License

---

Created with ❤️ by Cooper-X-Oak using Trae IDE