jnMetaCode/agency-agents-zh

agency‑agents‑zh – 中文版 AI 专家角色库

What it is

• A community‑maintained collection of 276 ready‑to‑use AI “agents” (expert‑level prompts) written in Chinese. Each agent encodes a full persona, workflow, and expected deliverables – e.g. a CEO, a security engineer, a TikTok strategist, a Qt/HMI developer, etc.
• The agents are not just simple system prompts; they describe how the expert thinks, what steps to follow, and what output to produce.
• The repository is a translation of the upstream agency‑agents project plus 63 original agents for the Chinese market (platform‑specific roles like 小红书运营, 国内网络工程师, 机械设计工程师, etc.).

Key features

Feature	Detail
Huge role library	276 agents covering 20 business departments (management, engineering, design, marketing, finance, GIS, security, etc.).
Chinese‑focused agents	63 roles created specifically for China‑centric platforms and industries.
Tool integration	Prompt files can be installed into 18 different AI‑coding assistants (Claude Code, Cursor, GitHub Copilot, OpenClaw, etc.) with a single script.
Zero‑code orchestration	Works with the companion Agency Orchestrator (CLI or native desktop app) to compose multi‑agent workflows, run them in parallel, and resume after failures.
Online browsing	All agents can be explored and copied from a web gallery (https://ao.aiolaola.com/experts).
Free educational material	Linked courses (33‑lesson “AI 专家团队实战”, 180‑lesson “从零学会 AI 编程”, 40‑lesson “从零构建 AI 智能体”).
Open source	MIT‑licensed, published on npm as agency-agents-zh.

How to get started

1. Install the agents into your preferred AI coding tool:

   # Detect installed tools and install automatically
   ./scripts/install.sh
   # Or target a specific tool, e.g. OpenClaw
   ./scripts/install.sh --tool openclaw
   

2. Run a task with the orchestrator (no code needed):

   npm install -g agency-orchestrator
   ao compose "帮我写一篇关于 AI Agent 的深度分析文章" --run
   

   The orchestrator will pick appropriate agents, let them collaborate, and return the final deliverable.
3. Use the desktop client (macOS/Windows/Linux) if you prefer a GUI – it bundles the orchestrator and the agent library without requiring Node.
4. Browse or copy prompts directly from the online gallery if you just need a single expert’s prompt.

Ecosystem & integrations

• AI coding assistants: Claude Code, Cursor, GitHub Copilot, OpenClaw, Kiro, Trae, OpenCode, Aider, Windsurf, Antigravity, Gemini‑CLI, Qwen‑Code, Codex, DeerFlow, WorkBuddy, CodeWhale, Hermes, Qoder, etc.
• Orchestrator: CLI (agency-orchestrator) and native desktop app for multi‑agent DAG execution.
• npm package: agency-agents-zh publishes the markdown files for programmatic consumption.
• Online UI: https://ao.aiolaola.com/experts provides searchable, filterable view of all agents.

Who should use it

• Developers who want to augment LLM‑based coding assistants with domain‑specific expertise.
• Product or marketing teams looking for AI‑generated strategies (e.g., TikTok growth plans, security audits, UI design specs).
• Educators and hobbyists who want a ready‑made library of expert prompts for learning or rapid prototyping.
• Companies targeting the Chinese market that need locally‑tuned AI agents for platforms like 小红书, 抖音, 微信, B站, etc.

License

• MIT License – free to use, modify, and redistribute.

---

All information above is taken directly from the repository’s README; no external assumptions have been added.