为持久 Agent 设计 Grok Bot Designing Grok Bot for a world of persistent agents
Designing Grok Bot for a world of persistent agents
从聊天会话改成名册:每个 Bot 有自己的电脑和例行。 From chat sessions to a roster, where each Bot has its own computer and routines.
Designing Grok Bot for a world of persistent agents
从聊天会话改成名册:每个 Bot 有自己的电脑和例行。 From chat sessions to a roster, where each Bot has its own computer and routines.
xAI 官方设计长文(2026-09-03):从聊天会话改成 Bot 名册、在场状态、Bot 自己的电脑,以及无需提示就能开工的 Routine。五件基本对象:Bot / Chat / Prompt·Skill·Routine / Tools / Artifacts。多 Agent 用幕僚长路由,记忆按角色边界。 Official xAI design essay (2026-09-03): from chat sessions to a Bot roster, presence, each Bot’s own computer, and Routines that start work without a prompt. Five primitives: Bots, Chats, Prompts/Skills/Routines, Tools, Artifacts. Multi-agent coordination via a Chief of Staff; memory stays role-scoped.
为什么推荐Why we recommend it 主源是 x.ai/news 官方设计文(Sep 3, 2026)。讲界面原语与持久 Agent,不是可直接开跑的角色配方;和 Guides / use-cases 目录互补。 Primary source is the official x.ai/news design essay (Sep 3, 2026). Covers interface primitives for persistent agents — complementary to Guides / use-cases, not a runnable role recipe.
文内谈委托与审批边界;对外发帖、付款、改生产仍要人审。 Essay discusses delegation and approval boundaries; still review before posting, paying, or changing production.
先在可撤回、不外发的范围内试跑。本条未批准前:文内谈委托与审批边界;对外发帖、付款、改生产仍要人审。。 Try it first in a reversible, non-outbound scope. Until approved: Essay discusses delegation and approval boundaries; still review before posting, paying, or changing production..
这条用法已包含可复用 Skill。跑通安全范围后,立刻把步骤固化下来。 This use already includes a reusable Skill. After the safe pass, lock the steps in.
输出稳定后再打开例行 / 定时。本条已有例行形态,不要一上来就无人值守外发。 Turn on the routine / schedule only after output is stable. This item already has a routine shape — do not start unattended outbound.