agency-agents-zh: a comprehensive library of 268 specialized AI agent roles with professional workflows for AI coding tools

What it solves

This project provides a comprehensive library of 268 specialized AI agent roles designed to move beyond generic prompts. It solves the problem of inconsistent AI output by providing agents with defined identities, professional workflows, and specific deliverables, allowing users to transform their AI programming tools into a team of virtual experts across 20 different departments.

How it works

The library consists of detailed Markdown files where each agent is defined by its persona, rules, and operational processes. These definitions can be integrated into various AI coding tools (like Claude Code, Cursor, or Copilot) via installation scripts or manual copying. For advanced users, the project suggests using the "Agency Orchestrator" to coordinate multiple agents to work together on complex tasks.

Who it’s for

Developers, marketers, and business professionals who use AI coding assistants or LLM-based tools and want high-quality, professional-grade outputs for specific roles such as security engineers, UI designers, or e-commerce operators.

Highlights

• Massive Role Library: 268 plug-and-play agents covering 20 departments including Engineering, Design, Marketing, Finance, and Legal.
• Localized for China: Includes 53 original agents specifically for the Chinese market, covering platforms like Xiaohongshu, Douyin, and WeChat.
• Broad Tool Support: Compatible with 18 different AI programming tools including Claude Code, Cursor, and Windsurf.
• Coded Workflows: Unlike simple prompts, these agents include specific thinking processes and delivery requirements (e.g., a security engineer following OWASP Top 10).
• Orchestration Support: Integration with a desktop client for zero-code agent orchestration and DAG parallel execution.