REA
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Claude skill that draws interactive isometric figures in a single HTML file
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/skills/mrbongoc-ai-iso-skill)在同一工作中把 AI 连到服务的 MCP
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Codebase intelligence for TypeScript and JavaScript, exposed as structured MCP analysis tools.
MCP server for terminal commands, file operations, and process management
同类工作,排序与 Skill 列表相同
| 本项 ai-iso-skill | rea | andrej-karpathy-skills | graphify | MoneyPrinterTurbo | |
|---|---|---|---|---|---|
| 作者 | MrBongoC | morluto | multica-ai | Graphify-Labs | harry0703 |
| Skill 数 | 1 | 1 | 1 | 1 | 1 |
| GitHub 星标 | 369 | 468,733 | 234,332 | 150,287 | 144,566 |
| 最近代码更新 | 2026年10月4日 | 2026年11月19日 | 2026年4月20日 | 2026年11月18日 | 2026年11月18日 |
| 许可证 | MIT | MIT | – | Apache-2.0 | MIT |
MrBongoC 提供的 1 个 Skill(指令集),AI 读取后照做 · 不是 MCP server
MCP = 手
把 AI 连到外部服务或数据,例如读取 Notion 或开发票。需要有程序或服务器在运行。
Skill = 手册
教 AI 怎么做某项工作的文本文件,不连接任何服务。
有些会重叠:少数 Skill 带有调用工具的脚本,少数 MCP 只提供知识。请从你想完成的工作出发。
Claude CodeClaude (เว็บ/แอป)
依据开发者 README 和 repo 中的文件夹
GitHub account MrBongoC (person), 5 years old
★ 369 stars, few users so far · code last changed 2026年10月4日
No, plain text files the AI reads and follows
No scripts, so it does not go online by itself
No hooks or writes to machine settings found
Checked automatically from the repo files at commit 44b4da7 (read only, nothing was run). Not a safety guarantee: read it and decide for yourself
第一步:你用哪个 AI?
选择你用的应用,本页只显示该应用的安装方法
Uploading to claude.ai would fail: the description is over 200 characters (332). Use "Let the AI install it" instead
复制这句话,粘贴到 Claude Code 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
重新打开 Claude Code,输入 /iso-figure,或直接提出相关任务,Claude 会自动选用
Checked against official docs 2026年11月12日 · Source: code.claude.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
正在检查命令 · 通过后才会显示复制按钮
摘自开发者 README,会下载并运行开发者的代码
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
ChatGPT 桌面应用内置 Codex,Codex 从 ~/.agents/skills 文件夹读取 Skill
在 ChatGPT 桌面应用中打开 Codex,把这句话粘贴到对话中
The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
把 iso-figure 文件夹放到 ~/.agents/skills/iso-figure
~ 是你的主文件夹($HOME)
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
不能使用你自己安装的 Skill,只能使用目录中插件自带的 Skill
想在 ChatGPT 中使用这个 Skill,请按上面的方法使用 ChatGPT 桌面应用
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
复制这句话,粘贴到 Codex 的对话中 · 适用于 ChatGPT 桌面应用中的 Codex、Codex CLI 和 IDE 扩展
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Codex 会自动发现新安装的 Skill;如果没有出现,请重启 Codex · 使用方法:在 ChatGPT 中输入 @ 并选择 Skill · 在 Codex CLI 或 IDE 扩展中运行 /skills,或输入 $ 加 Skill 名称
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
复制这句话,粘贴到 Cursor 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
在 Cursor 的 Agent 聊天中输入 / 并选择 iso-figure
Checked against official docs 2026年11月12日 · Source: cursor.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
正在检查命令 · 通过后才会显示复制按钮
适用于其他能读取 Skill 的应用。这句话列出了 Claude Code、Codex 和 Cursor 的路径,AI 会按你的应用选择 · Copy this sentence into the app’s chat
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Checked against official docs 2026年11月12日 · Source: code.claude.com, learn.chatgpt.com, cursor.com
其他支持 Agent Skills 的应用(如 Gemini CLI、GitHub Copilot)使用不同文件夹,请查看该应用的指南,再把含 SKILL.md 的文件夹复制过去。
已于 2026年11月11日 对照 Claude Code、Codex 和 Cursor 官方指南核实 skill 文件夹
以下文字来自开发者的 repo,不是 PopNx MCP 的指令
显示前 8,000 个字符 · 在 GitHub 打开完整文件
---
name: iso-figure
description: Build an interactive isometric line-art "figure" (Fig N style) of an object in a single HTML/SVG file: monochrome technical-illustration look, parts you can press that produce real output, live state readout. Use for hero illustrations, interactive diagrams, product figures, and playable desk objects (computer, synth, calculator).
---
# Interactive isometric figure
Produces one self-contained HTML file: a rounded panel captioned like a textbook plate, holding an isometric line drawing of an object whose parts actually work. A switch toggles, keys depress, a screen shows what was typed, a synth key makes sound. No libraries, no canvas, no 3D engine, just plain SVG generated by about 15 lines of projection code.
The object must do something real. A figure that only animates is decoration; a figure where pressing a part produces output (text, a number, a sound, a changed display) is the point. Decide what the output is before drawing anything.
## The four things that make it work
1. **Everything is a flat 2D shape placed on a 3D plane.** Never hand-compute skewed polygons. Draw ordinary `<rect>`, `<text>`, `<path>` in a face's local coordinates and let one SVG `matrix()` put it on the plane. Text, glyphs and rounded corners then sit in correct perspective for free.
2. **Objects are boxes; a box is its three visible faces.** Complex objects are boxes plus flat detail drawn on their faces (vents, slots, bezels, LEDs, labels).
3. **Interaction is class toggles on SVG groups.** One `state` object, one `render()`; CSS does the motion.
4. **The frame sells it.** Four monospace corner captions: `Fig N` (top-left), OBJECT NAME (top-right), the instruction (bottom-left), a live state readout (bottom-right).
## Kernel (copy verbatim, then build the scene)
```js
const C = Math.cos(Math.PI/6), S = Math.sin(Math.PI/6), OX = 465, OY = 300; // where the 3D origin lands in the viewBox
const P = (x,y,z) => [(x-y)*C + OX, (x+y)*S - z + OY]; // point: 3D -> screen
const D = (x,y,z) => [(x-y)*C, (x+y)*S - z]; // direction only
// plane = origin O + in-plane axes U,V. Anything inside a <g> with this transform lies flat on that plane.
const plane = (O,U,V) => { const o=P(...O), u=D(...U), v=D(...V);
return `matrix(${u[0]} ${u[1]} ${v[0]} ${v[1]} ${o[0]} ${o[1]})`; };
const TOP = (x,y,z) => plane([x,y,z],[1,0,0],[0,1,0]); // z = const
const FRONT = (x,y,z) => plane([x,y,z],[1,0,0],[0,0,-1]); // y = const (faces lower-left); pass its top-left corner
const SIDE = (x,y,z) => plane([x,y,z],[0,-1,0],[0,0,-1]); // x = const (faces lower-right); pass its top-left corner
const rect = (t,w,h,r=0,cls='face') =>
`<g transform="${t}"><rect class="${cls}" width="${w}" height="${h}" rx="${r}"/></g>`;
// box at (x,y,z), size w along x, d along y, h along z
const box = (x,y,z,w,d,h,r=0) =>
rect(SIDE(x+w,y+d,z+h), d,h,Math.min(r,h/4)) +
rect(FRONT(x,y+d,z+h), w,h,Math.min(r,h/4)) +
rect(TOP(x,y,z+h), w,d,r,'face top');
```
Axes: +x runs down-right, +y runs down-left, +z is up. The viewer sees the top, the `y = max` face (left) and the `x = max` face (right). Put the "front" of the object (screen, display) on a FRONT face and anything the user reaches for (keys, buttons, dials) at larger `y`, in front of it.
Detail on a face: open a group with that face's transform and draw in face-local units, origin at the face's top-left.
```js
svg += `<g transform="${FRONT(70,190,240)}">
<rect class="face glass" x="28" y="31" width="124" height="104" rx="12"/>
<text x="38" y="112" font-size="8.5">> hello</text>
</g>`;
```
## Required CSS
```css
.face { fill:var(--body); stroke:var(--line); stroke-width:1; vector-effect:non-scaling-stroke; stroke-linejoin:round; }
.face.top { fill:var(--deck); }
.detail { fill:none; stroke:var(--detail); stroke-width:1; vector-effect:non-scaling-stroke; }
.press { cursor:pointer; transition:transform 60ms ease-out; }
.press.down { transform:translateY(4px); }
.press.down .face { stroke:var(--live); }
```
`vector-effect:non-scaling-stroke` is not optional. Without it the matrix skews stroke widths and the hairline look is lost. Faces must have an opaque fill: painter's-order occlusion depends on it.
## Procedure
1. **Name the output.** What does pressing a part produce? Typed text, an arithmetic result, a note, a toggled light. Everything else serves that.
2. **Decompose the object** into 3-8 boxes. Write their `(x,y,z,w,d,h)` as a small table first. Pick the pressable parts and the one surface that displays state.
3. **Emit back to front.** Paint order is the only depth sorting: smaller `x+y` first, lower `z` first. For grids (keys, buttons) loop rows by increasing `y`, then columns by increasing `x`. Parts that sit on top of others (black piano keys over white) go after everything beneath them.
4. **Wrap each pressable part in its own `<g class="press" data-…>`** containing its box and its label. A label drawn on the part's TOP plane inside the same group moves with it when pressed.
5. **Frame it numerically.** Project the scene's extreme corners with `P`, then set `OX/OY` and `viewBox` for roughly 8-15% margin on every side. Eyeballing clips tall objects.
6. **Wire state.** One `state` object, one `render()` that toggles classes and rewrites the display and readout. Pointer and keyboard input call the same functions.
7. **Motion in CSS only.** Pressed part: `translateY(4px)` (straight down on screen is −z in iso) over about 60ms, stroke flashing to the live colour. Honour `prefers-reduced-motion`.
8. **Readout** in the bottom-right, terse and lowercase, dot-separated: `off` → `on · 9 chars · key t`. It changes on every interaction.
9. **Verify in a browser.** Load it headless, drive it with real key presses and mouse presses at projected coordinates, read back the readout, and screenshot one interacted state. Check nothing is clipped, nearer parts cover farther ones, and there are no console errors.
## Input patterns
Keyboard and pointer share one handler:
```js
const keys = new Map([...stage.querySelectorAll('.key')].map(k => [k.dataset.key, k]));
function press(k) {
const el = keys.get(k); if (!el) return false;
el.classList.add('down'); setTimeout(() => el.classList.remove('down'), 110);
/* …update state… */ render(); return true;
}
stage.addEventListener('click', e => { const k = e.target.closest('.key'); if (k) press(k.dataset.key); });
window.addEventListener('keydown', e => {
if (e.metaKey || e.ctrlKey || e.altKey) return;
if (press(e.key.length === 1 ? e.key.toLowerCase() : e.key)) e.preventDefault();
});
```
For parts that are held rather than tapped (instrument keys): use `pointerdown` / `pointermove` / `pointerup` on the svg with `setPointerCapture`, find the part under the pointer with `document.elementFromPoint(...).closest('.key')` so dragging across parts works, set `touch-action:none` on the svg, ignore `e.repeat` on keydown, and release everything on window `blur`.
For text typed onto a screen: show only the tail of the buffer, clip with a `<clipPath>` matching the glass rect, and escape `&` and `<` before writing into the SVG.
For sound: create the `AudioContext` inside the first press handler (browsers block it before a gesture), route through one master gain into an `AnalyserNode`, and draw the analyser's time-domain data as a `<polyline>` on the screen face each animation frame, starting from a rising zero crossing so the trace holds still.
## Look
- Define every colour as a CSS custom property and provide both a dark and a light set; the drawing works in either. Dark: panel `#141516` on page `#0c0d0e`. Light: panel `#f6f6f5` on page `#e9eaea`.
- Two greys only for linework (structure and detail). One near-white (near-black in light theme) `--live` colour reserved for whatever is live: lit display content, the pressed key, the selected button. That contrast is the whole effect; do not add hue, gradients or shadows.
- Two-tone parts (white and b提问、分享技巧或报告无法使用。按“有帮助”数量排序。