推荐在 Claude Code 中使用 Anatomy 技能
Highly recommend adding the Anatomy skill to Claude Code. Give it a technical co…
提供了非常具体的 Claude Code 技能安装步骤和完整的 Prompt 模板,展示了如何将 AI 能力应用于技术文档可视化,对开发者读者有直接的实操参考价值。
Highly recommend adding the Anatomy skill to Claude Code. Give it a technical concept and it 𝗯𝘂𝗶𝗹𝗱𝘀 𝗮 𝗺𝗮𝗰𝗵𝗶𝗻𝗲 𝘁𝗵𝗮𝘁 𝗮𝗰𝘁𝘀 𝗶𝘁 𝗼𝘂𝘁. The parts move with your mouse, and some come apart or turn. The launch post hit 1.9M views in two days, and the repo got 5,600 GitHub stars in four.
强烈建议为 Claude Code 添加 Anatomy 技能。给它一个技术概念,它就能 𝗯𝘂𝗶𝗹𝗱𝘀 𝗮 𝗺𝗮𝗰𝗵𝗶𝗻𝗲 𝘁𝗵𝗮𝘁 𝗮𝗰𝘁𝘀 𝗶𝘁 𝗼𝘂𝘁(构建出能演绎该概念的机器)。部件会随你的鼠标移动,有些还能拆解或转动。发布帖在两天内获得了 190 万次观看,仓库在四天内获得了 5,600 个 GitHub Star。
It first invents a machine whose motion follows the concept, then builds it out of dozens of small shaded parts. → A cache that skips the work when nothing changed becomes 𝗮 𝗽𝗶𝗻-𝘁𝘂𝗺𝗯𝗹𝗲𝗿 𝗹𝗼𝗰𝗸 𝘁𝗵𝗮𝘁 𝗼𝗻𝗹𝘆 𝘁𝘂𝗿𝗻𝘀 𝗼𝗻𝗰𝗲 𝗲𝘃𝗲𝗿𝘆 𝗽𝗶𝗻 𝗹𝗶𝗻𝗲𝘀 𝘂𝗽 → One GPU context serving many canvases becomes a gantry plotter whose single print head visits 16 wells in turn
它首先发明一种运动符合该概念的机器,然后用数十个小型着色部件将其组装出来。 → 当无变化时跳过工作的缓存变成了 𝗮 𝗽𝗶𝗻-𝘁𝘂𝗺𝗯𝗹𝗲𝗿 𝗹𝗼𝗰𝗸 𝘁𝗵𝗮𝘁 𝗼𝗻𝗹𝘆 𝘁𝘂𝗿𝗻𝘀 𝗼𝗻𝗰𝗲 𝗲𝘃𝗲𝗿𝘆 𝗽𝗶𝗻 𝗹𝗶𝗻𝗲𝘀 𝘂𝗽(只有当所有销钉对齐时才转动一次的 pin-tumbler lock/弹子锁) → 服务于多个画布的单个 GPU 上下文变成了一个龙门绘图仪,其唯一的打印头依次访问 16 个井位
You get a standalone HTML page, an SVG or React components. Put the interactive version in your docs or on your site, and a screenshot in the README. Fire and water are drawn with WebGL. It zooms in to check every joint, so a single figure usually takes 30 to 90 minutes.
你会得到一个独立的 HTML 页面、SVG 或 React 组件。将交互版本放入你的文档或网站中,并在 README 里放一张截图。火和水使用 WebGL 绘制。它会放大以检查每个关节,因此单个图形通常需要 30 到 90 分钟。
Run npx skills add wheresryan22/anatomy -g -a claude-code in your terminal (it needs Node 22.20 or later). No API key required. Then open a new Claude Code session and send this to Opus 5.5 👇
在你的终端中运行 npx skills add wheresryan22/anatomy -g -a claude-code(需要 Node 22.20 或更高版本)。无需 API 密钥。然后打开一个新的 Claude Code 会话,并将以下内容发送给 Opus 5.5 👇
"/anatomy Use a working machine to explain the hardest-to-explain mechanism in this project. Make it interactive, light theme. Read the code first and give me 3 candidate mechanisms to choose from, each with one line on what machine you'd build. Start only after I pick. Take every input, output and readout in the figure from the real numbers in the code. Don't make anything up. When it's done, export a standalone HTML file into docs/, then use the skill's own scripts/capture.mjs to take a screenshot and put it in the README."
"/anatomy 用一台正在运行的机器来解释本项目中最难解释的机制。使其具有交互性,采用浅色主题。 先阅读代码,给我 3 个候选机制供选择,每个机制用一行说明你会构建什么样的机器。在我选定之前不要开始。 从代码中的实际数值中提取图中所有的输入、输出和读数。不要捏造任何内容。完成后,将独立的 HTML 文件导出到 docs/ 目录,然后使用该技能的 scripts/capture.mjs 脚本截取一张截图并放入 README 中。"
https://github.com/wheresryan22/anatomy
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力