jnMetaCode/agency-agents-zh

⭐ 10,347  ·  #4  ·  Shell

🎭 211 plug-and-play AI expert roles — supporting 16 tools including Hermes Agent/Claude Code/Cursor/Copilot, covering 18 departments such as engineering/design/marketing/finance. Includes 46 original Chinese market agents (Xiaohongshu/Douyin/WeChat/Feishu/DingTalk, etc.)

Shell agency-orchestrator agent-definitions ai-agents Skill

Project Analysis

🎯 Positioning	Agent capability enhancement
💡 Core Value	Provides standardized Skills and Prompt templates for AI coding Agents, covering specific scenarios (code review, debugging, architecture design, etc.), enabling higher quality output in these scenarios
👥 Who It's For	Developers using Agent tools like Claude Code/Cursor/Codex who want to improve Agent performance on specific tasks

Why It's Worth Attention

10,347 Stars, good community activity, indicating it solves real pain points. Developed using Shell.

AI Deep Analysis Report

One-Sentence Summary

211 Chinese AI expert role presets, plug-and-play.

Core Features

This project is essentially a curated collection of AI role definitions and system prompts, rather than a traditional runnable framework. Its core value lies in the breadth and practicality of its content.

1. Massive Ready-to-Use Role Library: Provides 211 predefined AI expert roles covering 18 virtual departments such as engineering, design, marketing, finance, and human resources. Each role includes detailed system prompts defining its behavior, knowledge, and output style.
2. Multi-Tool Compatibility: Explicitly supports 16 mainstream AI tools and platforms, including Hermes Agent, Claude Code, Cursor, GitHub Copilot, DeepSeek, etc. This is achieved by providing configuration files (e.g., cursorrules, claude.md) tailored for different tools, ensuring role definitions are correctly loaded and parsed.
3. Localized Agents: A major highlight is the inclusion of 46 AI agents customized for popular Chinese market platforms (e.g., Xiaohongshu, Douyin, WeChat, Feishu, DingTalk). These agents understand the Chinese internet ecosystem, marketing language, and operational logic, offering immense value to local developers.
4. Structured Organization: Roles are clearly organized by "department" and "tool" in the directory structure, with index files like agents.json provided for easy batch searching and loading via scripts or tools. This structured design is key to the project's quality as a "content collection."

Technical Architecture

• Primary Tech Stack: Shell. The project itself contains no core code to compile or run; all content consists of .md, .txt, .json, and .sh files. Shell scripts are primarily used for automated installation and role file distribution.
• Code Structure Highlights:
  • Clear Directory Hierarchy: Organized under /agents/ as Department -> Role -> Tool. For example, the path /agents/marketing/social-media-manager/cursorrules clearly indicates the role, its department, and the applicable tool.
  • Configuration-Driven: Provides a role manifest via agents.json, allowing other tools or scripts to programmatically access all role information. This is a low-coupling, high-maintainability design.
  • One-Click Installation Scripts: Scripts like install.sh encapsulate the complex process of copying role files to the corresponding tool configuration directories (e.g., Cursor's .cursor/rules), lowering the barrier for users.

Quick Start Guide

This project doesn't need to be "run"; the core is "application."

1. Clone the Repository:

   git clone https://github.com/jnMetaCode/agency-agents-zh.git
   cd agency-agents-zh

2. Select Role and Tool: Navigate to the /agents/ directory, find your desired role (e.g., marketing/social-media-manager), then choose your tool (e.g., cursorrules).

3. Apply the Role:

   • Method 1 (Manual): Copy the content of the configuration file (e.g., the cursorrules file) from the role's folder to your project's .cursor/rules/ directory, or paste it into the system prompt input box of your AI tool.
   • Method 2 (Automatic): Run the project's installation script (if it exists and supports your tool). For example, for Cursor, you could run bash install.sh cursor to install all Cursor-compatible roles at once.

Strengths, Weaknesses, and Use Cases

Strengths:

• High Content Value: Provides a large number of well-designed, ready-to-use Chinese role prompts, with the localized agents filling a market gap.
• Good Developer Experience: Structured organization, multi-tool adaptation, and one-click installation scripts significantly lower the usage barrier, allowing non-prompt engineers to easily leverage advanced roles.
• High Practicality: Plug-and-play, no coding required, can directly improve the performance of daily development tools like Cursor and Copilot on specific tasks.

Weaknesses:

• Not a Framework, No Runtime: The project itself does not provide complex capabilities like role orchestration, dialogue management, or multi-agent collaboration; it is merely a prompt repository.
• Quality Depends on Maintenance: The quality and timeliness of role prompts rely entirely on the maintainer's continuous updates and community contributions. Some roles may be ineffective or outdated.
• Limited Installation Methods: Currently relies mainly on file copying. Integration can be difficult for tools that do not support custom system prompts or rules.

Use Cases:

• AI Application Developers: Teams needing to quickly prototype AI assistants for specific tasks (e.g., generating Xiaohongshu copy, writing technical documentation, performing code reviews).
• Developers Using AI Coding Assistants: Those looking to improve the professionalism and output quality of Cursor, Copilot, Claude Code in specific domains (e.g., frontend development, backend architecture).
• Enterprise AI Deployment Teams: Teams needing to quickly deploy standardized AI work assistants for different departments (e.g., marketing, product, customer service), reducing the management cost of prompt engineering.

Community and Popularity

• Stars (10,347): Very High. For a "content collection" project primarily using Shell, this Star count fully demonstrates the popularity and market demand for its content, especially within the Chinese developer community.
• Topics: Covers multiple hot areas like ai-agents, prompt-engineering, chinese, cursor-rules, claude-code, reflecting its broad applicability.
• Last Update (2026-05-09): The project is actively maintained, with recent updates indicating the maintainer is keeping up with tool and role changes, a positive sign.
• Fork Trend: Although not directly provided, a high Star count is usually accompanied by a significant Fork count, indicating community demand for secondary development and customization.

Summary: jnMetaCode/agency-agents-zh is a high-quality, high-value AI role prompt resource library. Its success stems not from technological innovation but from precisely addressing the "Prompt Engineering" pain point of Chinese AI users. For any Chinese user or developer looking to quickly and efficiently use AI tools for specific tasks, this is a treasure trove project worth bookmarking and using.

Technical Information

• 💻 Language: Shell
• 📂 Topics: agency-orchestrator, agent-definitions, ai-agents, ai-roles, chinese
• 🕐 Updated: 2026-04-23
• 🔗 Visit GitHub Repository

---

Data updated on 2026-05-09 · Stars count based on actual GitHub data