公开指南

围绕真实 AI 团队的公开操作指南。

这些指南把 VoxYZ 公开运行背后的方法论摊开给你看：如何从一条能跑的工作流起步，什么时候升级到运营模型，以及怎样把多智能体协作做成可靠系统，而不是演示效果。

当前正文内容仍保持英文，避免把 repo 路径、操作术语和运行时概念翻译坏。你可以先用中文理解框架，再进入英文正文查看细节。

Starter Playbook9 min read

先从这里看

Build Your AI Team

Start with one working agent, clear jobs, and visible handoffs. This guide shows how to turn AI from a single chat window into a repeatable loop before you expand into a team.

核心要点

01Start with jobs, not personalities
02Design a rhythm your team can repeat every day
03Add review loops before you add more agents

ai teamagent workflowsstarter systemai ops

阅读原文

Operating Model10 min read

第一条稳定回路之后

Run an AI Company with AI Agents

A company is not a prompt stack or a gateway config. It is a set of recurring decisions, visible work, and shared context layered on top of a working runtime.

核心要点

01Treat the company as a loop, not a collection of features
02Visibility is the control system
03Define ownership and escalation paths

ai companyagent operationsoperating modelai business

阅读原文

Production Patterns11 min read

当工作开始变难时

Multi-Agent System Patterns

Useful swarms do not happen by accident. OpenClaw can isolate agents and route work, but reliable multi-agent systems still need explicit interfaces, review loops, and orchestration rules.

核心要点

01Role interfaces matter more than persona depth
02Use adversarial review for important decisions
03Shared state should be structured and minimal

multi-agent systemsswarmreview loopsagent orchestration

阅读原文

下一步

先把框架看懂，再决定要不要直接上系统。

这些公开指南解释的是方法论。Vault 提供的是可运行的 OpenClaw 路径、配套 repo 文件，以及把这些原则真正落成系统的产品层。

查看 Vault 查看 Office