Trae AI Agent System

English | 中文

About This Project

Trae AI Agent System is an AI Agent enhancement system for Trae IDE. It provides Rule-based routing, professional Skills, and persistent Memory capabilities — transforming your Agent from "can do anything but unstable" to "fast when simple, rigorous when complex, always learning".

Architecture

User Input → Rules (Routing & Constraints) → Skills (Execution) → Memory (Learning)

| Layer | Count | Responsibility | |:------|:------|:---------------| | Rules | 8 | Routing decisions, behavior constraints, environment handling | | Skills | 33 | Professional toolboxes covering design → coding → debugging → commit → completion | | Memory | Core Memory | Cross-session knowledge accumulation |

Core Features

• T-Shirt Sizing: Automatic task classification (S/M/L) — small tasks are fast, large tasks are thorough
• Closed-Loop Quality: Every skill produces verifiable evidence — no "should be fine" allowed
• Self-Improvement: Lessons learned are stored in Core Memory — never repeat the same mistake
• Chinese Team Ready: Native Chinese triggers, domestic Git platform support, bilingual skill files
• Windows Native: PowerShell commands, port conflict recovery, path conventions

Quick Start

Method 1: npm install (Recommended)

Run this anywhere (installs to your home directory, global for all projects):

npx trae-agent-enhancements

Follow the interactive prompt to select your Trae edition:

| Edition | Install Location | Description | |:--------|:-----------------|:------------| | China (trae.cn) | ~/.trae-cn/rules/ + ~/.trae-cn/skills/ | Global, applies to all projects | | International (trae.ai) | ~/.trae/rules/ + ~/.trae/skills/ | Global, applies to all projects |

You can also specify the edition directly with --edition:

npx trae-agent-enhancements --edition cn       # China edition
npx trae-agent-enhancements --edition intl     # International edition
npx trae-agent-enhancements --help             # View help

Prerequisites (MCP)

Some skills require the following MCP Servers for full functionality:

| MCP Server | Purpose | Installation | |:-----------|:--------|:-------------| | Everything Search | Windows local file search | Add in Trae Settings → MCP, see everything-search skill for config | | Chrome DevTools MCP | Browser automation, Console/Network/DOM debugging, performance analysis | Add npx chrome-devtools-mcp in Trae Settings → MCP |

These MCPs power the everything-search and chrome-devtools skills. Without them, related skills will not work.

Method 2: Manual Install

# Clone the repository
git clone https://github.com/MorningStar0709/trae-agent-enhancements.git

# Choose your edition and copy to the corresponding directory

# China edition (trae.cn)
cp -r trae-agent-enhancements/rules  ~/.trae-cn/rules
cp -r trae-agent-enhancements/skills ~/.trae-cn/skills

# International edition (trae.ai)
cp -r trae-agent-enhancements/rules  ~/.trae/rules
cp -r trae-agent-enhancements/skills ~/.trae/skills

Try These Commands

"帮我排查这个报错" (Help me debug this error) "先写计划再实现" (Write a plan first, then implement) "改好了没？验证一下" (Is it fixed? Verify it) "帮我提交" (Help me commit) "记住这个处理方式" (Remember this approach)

Documentation

| Doc | Description | |:----|:------------| | docs/01-intro.md | 15-second overview | | docs/02-overview.md | Features & highlights (3 min) | | docs/03-components.md | Component quick reference (5 min) | | docs/04-design.md | Design decisions & rationale (5 min) | | docs/05-architecture.md | Complete architecture & workflows (15 min) |

Skill Paths

| Path | Skills | |:-----|:-------| | Design & Planning | brainstorming → writing-plans → executing-plans / subagent-driven-development | | Debugging & Quality | systematic-debugging → test-driven-development → verification-before-completion | | Completion & Evolution | git-commit → finishing-a-development-branch → self-improvement | | Orchestration | dispatching-parallel-agents, workflow-runner, find-docs | | Browser & Frontend | chrome-devtools, frontend-design, chart-visualization, a11y-debugging | | Meta Skills | skill-creator, skill-stability-review, skill-language-policy, creating-trae-rules |

Windows/Trae Adaptation

• PowerShell Commands: Primary syntax for Windows environments
• Port Recovery: netstat → taskkill → verify → retry
• Path Conventions: Forward slashes in globs, absolute paths with backslashes
• Core Memory: 20-entry limit per scope, auto-eviction for stale entries

Acknowledgments

This project references superpowers-zh (AI 编程超能力 · 中文增强版) for its npx installation pattern and project structure design. Thanks to @jnMetaCode for the excellent work.

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines.

For Chinese version, see CONTRIBUTING_zh.md.

License

This project is licensed under the MIT License.