Grok Bot工作计划AI 评分 · 72/100AI 辅助编辑·未发现明显提示词注入风险X

用 Grok Bot 搭建你的 AI 操作系统

输入:Grok Bot 舰队 → 处理:审计清理、分级指挥、实时城市、决策队列 → 输出:可管理的 AI 工作流与单一入口

有一点经验步骤流程贴士
给谁
拥有多个 Grok Bot 且需要统一管理入口的用户
你会得到
从 79 个 Bot 精简到 30 个,建立单一首席助手(Cooper)入口,用像素城市仪表盘监控状态与待办
你需要
Grok Bot

硬约束

需要 Grok Bot 账号及多个已创建的 Bot;城市构建提示词需粘贴至主 Grok Bot;图标绘制需 Python + Pillow(可选);来源未说明费用条件。

他们做了什么

作者 JordanRafealov 曾创建了 79 个 Grok Bot,发现其中 37 个从未被使用。他进行舰队审计,按「是否有明确职责」和「是否有人使用」评分,将 Bot 从 79 个精简到 30 个。随后建立了以猫 Bot Cooper 为首席助手的四级公司架构,设定 6 条规则确保工作向上流转、决策只上报给他。他用一段长提示词让 Grok Bot 生成了一个实时像素艺术城市仪表盘,展示所有 Bot 的状态、目标进度和待审批队列。最后设置了决策仪表盘,只显示需要他审批的事项,并定义了审批边界(发送、发布、花钱、部署)。

效果如何

作者报告:审计后舰队能自我清理(每周五自动运行);工作不再由他追逐,而是自动上报等待确认;像素城市让他每天主动打开查看;决策仪表盘解决了他作为瓶颈的问题。作者自述此为 author_report,非本站实测。

值得知道

审计有效的前提是诚实面对不用的 Bot;组织架构有效的前提是每个部门有真实数字目标;城市仪表盘有用是因为作者确实每天打开;审批流仍需人类参与,作者故意保留最终签字权。其余提示词(审计、每周清理 skill、组织架构、Bot 图标、决策仪表盘)作者称在其 Skool 社群中提供,未包含在本来源内。Grok Bot Pages 功能的可用性与性能限制未说明。Bot 图标绘制需要 Python + Pillow 代码能力。

步骤

  1. 舰队审计:评估每个 Bot 的价值

    给每个 Bot 打分(1–5 分),回答两个问题:它是否有明确的单一职责?是否确实有人在用?Grok Bot 会自动审计整个舰队并返回每个 Bot 的裁决。作者的结果:27 保留、11 合并、4 搁置、37 删除。审计还会发现无人负责的工作缺口,可据此新建 Bot 填补。将审计保存为每周五自动运行的 skill。

  2. 建立组织架构图与指挥链

    模仿真实公司结构:你位于顶层,下设一位首席助手(作者用名为 Cooper 的猫 Bot),再下设各部门负责人(如 YouTube、财务、 agency),每位负责人对应一个具体数字目标。负责人之下为执行 Bot(SEO、QA、客服等)。设定 6 条规则:只与一位 Bot 对话;工作向上汇报;只把决策提交给你;用文件而非闲聊;4 级架构,部分 Bot 可兼岗;每个部门一个目标。任务流转:你丢出待办清单 → Cooper 指派负责人 → 负责人分配给执行 Bot → QA 检查 → 返回给你最终确认。

  3. 用提示词构建像素城市仪表盘

    在主 Grok Bot 中粘贴下方提示词,生成一个实时像素艺术城市页面(Grok Bot Page 或单文件网页应用)。城市以等距视角展示各部门区域,每个 Bot 有一张桌子,状态通过屏幕发光/粉色感叹号/暗淡显示。点击桌子可查看 Bot 卡片。顶部显示 CEO 等级、XP、各部门目标进度条。左侧面板列出所有 Bot 及其状态,右侧面板「等你决策」按优先级排列待审批事项。底部有实时完成的滚动 ticker。支持 REC(录屏隐私遮罩)和 SHOWCASE(演示数据)模式。

    运行提示词 · 来源原文。主 Grok Bot 对话输入框,生成 Grok Bot Page 或单文件网页应用

    Build me "My AI Operating System", a live pixel-art city of my AI company, as a Grok Bot Page. If a Page can't run it smoothly, build it as a single-file web app instead. Use my real bots and real data only. Never make up a number. If you can't see something, show "unknown".
    
    DATA EACH BOT REPORTS
    At the end of every job it already runs, each bot saves one status line: name, division, department, state (working, waiting on me, idle), what it's doing now in one line, what it finished, what it needs from me, its number today against its daily target, last active. No extra runs just to report. The operating system reads the newest status lines when I open it and refreshes them every few minutes while it's open.
    
    THE CITY
    - An isometric pixel-art office. Each division is its own zone with its name in neon and a count like "19 bots, 3 need you".
    - Every bot is a small pixel robot at a desk in its zone. My Chief of Staff sits in the middle as a cat.
    - State on every desk: glowing screen = working, pink "!" = waiting on me, dim = idle. Add a small legend the first time I open it.
    - Click a bot to open its card: job, what it's doing now, what it needs from me, its number, last active, and a button to open its chat.
    
    AROUND THE CITY
    1. Top bar: my CEO level and XP, one goal bar per division (the number, the target and the date, filled forward as we get closer, on pace or at risk), the local time, and buttons: SND, MUSIC, REC, SHOWCASE.
    2. Tabs under the top bar: All, then one per division. A tab zooms into that zone.
    3. Left panel "Your bots": every bot grouped by division, one line on what it's working on, a colored dot for its state, and a count of working, waiting and idle.
    4. Right panel "Waiting on you": every decision my bots need, ranked by deadline, money and age. Same ask from 3 bots? Show it once with "3 bots". Each row gets Approve, Decline, Answer or Go to bot.
    5. A daily quest card above the queue, for example "Get the queue to 3 or fewer".
    6. A ticker at the bottom with the latest finished work.
    
    APPROVALS AND THE RELAY
    - I only approve things that send, post, spend money or deploy. Everything else gets done without me.
    - Approve or Decline here never sends anything itself. It saves my answer for my Chief of Staff, who passes it to the right bot and marks the item done.
    - Clearing items gives me XP. Keep score in this browser.
    
    REC AND SHOWCASE
    - REC button: masks every $ amount, client name and email so I can film my screen. Remember the setting.
    - SHOWCASE button: swaps everything for clearly fake demo data: healthy goal bars, round numbers, generic clients like "Client A", busy bots. Show a "SHOWCASE DATA" tag next to the title. Off by default. Showcase numbers are fake, so they show even with REC on.
    
    LOOK
    Dark night city, neon purple, cyan, orange and pink, rain and blinking windows, pixel fonts. Optional soft music and click sounds, off by default. Keyboard: 1 to 6 for tabs, R for REC, M for sound, L for music.
    Your bots save one status line at the end of jobs they already run, and the city reads those lines. No extra runs just to report.
  4. 设置决策仪表盘与审批流

    建立一个只显示「需要你」事项的仪表盘:各部门目标、当前状态、优先级决策队列、审批/完成/保持开放按钮、实时活动日志。作者设置了一个从 Command 到 Content 到 Autonomous 到 Dopamine 的旋钮(装饰性)。关键规则:只审批发送、发布、花钱或部署的事项,其余自动执行。审批/拒绝仅保存你的答复,由首席助手传递给对应 Bot 执行。将仪表盘放在副屏或 Grok Bot Pages 中常驻打开,定期清理队列。

可复制提示词

参考提示词 · 本站建议版本。使用前确认粘贴位置、输入变量及输出格式。

这一段是提示词,页面不会在这里插入广告。

无提示词,按步骤操作即可。

来源

Turn Grok Bot Into Your AI OS:fleet audit → org chart → office city → decisions · JordanRafealov · X · 2026-10-30T12:57:37.000Z

留言与反馈

试过这个方法吗?分享效果、遇到的问题或改进建议。留言审核后公开,请勿填写敏感信息。

正在加载…

    相关用例