Developers Digest: Routines invert automation (do the job, then schedule)
先把活做完再排期，而不是先搭触发器。 Finish the job, then schedule it, instead of building the trigger first.
2026-08-23 长文：对照 Zapier/n8n「先搭触发器」与 Grok Bot「先做完再存 Skill / Routine」。写清 Teach a task 草稿边界、六要素排期（Bot/时区/输入/输出/审批/缺数策略）、适合低判断多步骤琐事，以及演示型 Routine 不如节点工作流可审计的诚实代价。 Aug 23, 2026 essay: contrasts Zapier/n8n “build triggers first” with Grok Bot’s “finish the job, then save a Skill/Routine.” Covers Teach-a-task drafts, the six scheduling facts (Bot / timezone / inputs / outputs / approval / missing-data policy), fit for low-judgment multi-step chores, and the honest tradeoff that demonstrated routines are less auditable than node graphs.
为什么推荐Why we recommend it 第三方解读（公开页 200，2026-09-17）。机制描述对齐官方文档；产品发布日期与措辞以 xAI 原文为准。 Third-party explainer (public page 200, 2026-09-17). Mechanics track official docs; launch dates and wording should follow xAI’s own posts.
外发、购买、删除、发布与改生产系统必须审批；缺源时报告而非用旧数据硬跑。 Sending, purchasing, deleting, publishing, and production writes stay behind approval; missing sources should be reported, not silently filled with stale data.
先在可撤回、不外发的范围内试跑。本条未批准前：外发、购买、删除、发布与改生产系统必须审批；缺源时报告而非用旧数据硬跑。。 Try it first in a reversible, non-outbound scope. Until approved: Sending, purchasing, deleting, publishing, and production writes stay behind approval; missing sources should be reported, not silently filled with stale data..
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.
