REA
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Teach your AI to draw correct, beautiful draw.io diagrams — declarative layout engine, ground-truth stencils, structural validator, vision self-check. AWS · Azure · GCP · Databricks · BPMN. Zero dependencies.
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/sparklabx-drawio-ai-kit)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 drawio-ai-kit | rea | andrej-karpathy-skills | graphify | MoneyPrinterTurbo | |
|---|---|---|---|---|---|
| 作者 | sparklabx | morluto | multica-ai | Graphify-Labs | harry0703 |
| Skill 数 | 1 | 1 | 1 | 1 | 1 |
| GitHub 星标 | 674 | 467,235 | 234,125 | 149,873 | 144,332 |
| 最近代码更新 | 2026年10月12日 | 2026年11月18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 |
sparklabx 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 CodeCodexGemini CLI
依据开发者 README 和 repo 中的文件夹
GitHub account sparklabx (organization) is only 5 months old: take extra care
★ 674 stars · code last changed 2026年10月12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 8f983cd (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 (397). 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 /drawio,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。
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 drawio 文件夹放到 ~/.agents/skills/drawio
~ 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 drawio
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
5,152 个字符 · 在 GitHub 打开完整文件
---
name: drawio
version: 2.0.0
description: Use when the user asks for any draw.io diagram — cloud architecture (AWS, Azure, GCP, Databricks, multi-cloud, hybrid/DR, landing zone, VPC/network, serverless, data pipeline) or a business process (BPMN, swimlane, workflow with roles). Builds with the drawio-ai layout engine using real stencils, validates, and runs a render self-check. Default output is a .drawio file; PNG/SVG only on request.
license: MIT
---
# draw.io diagrams
You build the diagram by writing a small Node script that uses the `drawio-ai` engine.
The engine places every box and routes every arrow. You only declare *what* is inside *what*.
Follow the steps in order. Do not skip a step. Each step names the exact file to read —
read only that file, nothing else in the kit.
## Step 0 — Check the CLI
```bash
command -v drawio-ai >/dev/null 2>&1 && drawio-ai root || echo "MISSING"
```
If it prints `MISSING`: stop. Tell the user to install it, then try again:
```bash
npm i -g drawio-ai-kit
```
Never run `npm i -g` yourself.
## Step 1 — Understand the request
You need these 4 facts. Take them from the user's message first.
| Fact | Example | If missing |
|---|---|---|
| **Domain** | AWS · Azure · GCP · Databricks · multi-cloud · BPMN | Ask. |
| **What to draw** | "3-tier web app in one VPC, 2 AZs" · "order approval process" | Ask. |
| **Output path** | `./docs/arch.drawio` | Use `<cwd>/<short-name>.drawio`. Do not ask. The build script goes in the same folder. |
| **Format** | `.drawio` (default) · also `.png` | Use `.drawio`. Do not ask. |
Rules for asking:
- Ask **one** short message with all missing questions together. Offer choices, e.g.
"Which cloud: AWS, Azure, or GCP?"
- If the user said "just do it" or the request is already clear, do not ask. Pick the
common choice and tell the user what you assumed at the end.
- The source is a Terraform/Terramate repo? Follow `workflows/from-iac.md` before Step 2.
## Step 2 — Pick the domain reference
Read **exactly one** row's files from this skill folder:
| Domain | Read | Template to start from (`drawio-ai scaffold --list` shows all) |
|---|---|---|
| AWS | `references/aws-architecture.md` | `build_vpc.mjs`, `build_serverless.mjs`, `build_pipeline.mjs`, `build_landingzone.mjs` |
| Azure | `references/azure-architecture.md` | `build_azure_vnet.mjs`, `build_azure_hub_spoke_lz.mjs` |
| GCP | `references/gcp-architecture.md` | `build_gcp_vpc.mjs`, `build_gcp_shared_vpc_landing_zone.mjs` |
| Databricks | `references/databricks-architecture.md` | `build_lakehouse.mjs`, `build_data_platform.mjs` |
| Multi-cloud / hybrid | the reference of each cloud involved | `build_multicloud.mjs`, `build_hybrid.mjs` |
| BPMN / swimlane | `references/bpmn.md` | `build_bpmn.mjs` |
Then also read:
- `references/api.md` — the whole engine API on one page (every domain). Never open engine source code.
- `references/principles.md` — layout rules: dense grids, flow direction, no overlaps (cloud domains; skip for BPMN).
Optional, only if you are unsure which layout fits: `references/diagram-types.md`.
Colors and creators: `references/style-guide.md`.
## Step 3 — Build, check, fix
Follow `workflows/build.md`. Short version:
1. `drawio-ai search "a, b, c"` — look up **all** icons in one call (comma or space separated, e.g. `"k8s pg es"`). Use only names it returns.
2. `drawio-ai scaffold <template>.mjs -o <out-dir>/build.mjs --name <name>.drawio` — copy the closest
template. `<out-dir>` = the folder of the output path from Step 1; the `.drawio` lands there.
3. Edit the tree in `build.mjs`. No x/y numbers.
4. `drawio-ai run <out-dir>/build.mjs` — it builds, prints `VALIDATE: {…}`, and renders a PNG with an `issues` list.
5. Fix **all** errors/warnings/advice/issues in one edit. Run again. Repeat until all are empty (max 5 runs).
6. Look at the PNG once. Fix what you see. Done. (No draw.io desktop app → no PNG: skip this step.)
Can your harness run a subagent that has shell + image reading? You may hand Step 3
to it with `workflows/delegate.md`. Otherwise do Step 3 yourself.
## Step 4 — Deliver
Check every box before you answer:
- [ ] Built with the layout engine (`group`/`frame`/`grid` + `icon`/`box` + `renderTree`). No x/y numbers.
- [ ] Every icon name came from `drawio-ai search`. No icon recolored.
- [ ] `drawio-ai validate <file>.drawio` → `"ok":true` with empty `errors`, `warnings`, `advice`.
- [ ] You looked at the rendered PNG — or, if `RENDER-SKIPPED` was printed, you told the user no PNG was made.
- [ ] The file is in the user's project — never inside the kit folder (`drawio-ai root`).
Reply with: the `.drawio` path, the `.png` path (if any), one sentence on what the
diagram shows, and any assumptions you made.
## Never
- Never write files into the kit folder (`drawio-ai root`).
- Never invent icon names or style strings. Search for them.
- Never hand-write draw.io XML or x/y coordinates.
- Never read the engine's source (`dist/`, `src/`). `references/api.md` has everything.
- Never import by file path. Scripts use `import { ... } from "drawio-ai-kit"` and run with `drawio-ai run build.mjs`.
提问、分享技巧或报告无法使用。按“有帮助”数量排序。