REA
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Agent-native PowerPoint manipulation — one CLI lets AI agents inspect, edit, create, and verify .pptx files through atomic JSON patches. Packaged as an Agent Skill.
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/everyinc-hands-on-deck)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 hands-on-deck | rea | andrej-karpathy-skills | graphify | MoneyPrinterTurbo | |
|---|---|---|---|---|---|
| 作者 | EveryInc | morluto | multica-ai | Graphify-Labs | harry0703 |
| Skill 数 | 1 | 1 | 1 | 1 | 1 |
| GitHub 星标 | 224 | 465,460 | 233,910 | 149,364 | 144,149 |
| 最近代码更新 | 2026年8月4日 | 2026年11月17日 | 2026年4月20日 | 2026年11月17日 | 2026年11月17日 |
| 许可证 | MIT | MIT | – | Apache-2.0 | MIT |
EveryInc 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)Codex
依据开发者 README 和 repo 中的文件夹
GitHub account EveryInc (organization), 5 years old
★ 224 stars, few users so far · code last changed 2026年8月4日
7 scripts that can run on a computer: scripts/inventory.py, scripts/rearrange.py, scripts/replace.py, scripts/thumbnail.py and 3 more
No internet calls found in the scripts
No hooks or writes to machine settings found
Checked automatically from the repo files at commit a24b996 (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 (448). 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 /hands-on-deck,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 hands-on-deck 文件夹放到 ~/.agents/skills/hands-on-deck
~ 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 hands-on-deck
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: hands-on-deck
description: "Use this skill any time a .pptx file is involved in any way — as input, output, or both. This includes reading, analyzing, or extracting content from presentations; editing text, images, styles, or layout of existing decks; creating new slides, shapes, tables, or pictures; reordering, duplicating, or merging slides across decks; and verifying decks visually. Trigger whenever the user mentions a deck, slides, a presentation, or a .pptx filename."
---
# hands-on-deck — agent-native PPTX manipulation
One tool does everything: `scripts/deck.py`. You write a JSON *patch* describing your edits; deck.py validates and executes it atomically, then lints the result. You never need to open slide XML for routine work.
**Golden rules** (the tool enforces and repeats these):
- ALL slide indices are 0-based, everywhere. Shapes are addressed by stable native ids (`s12`) or unique shape name.
- Positions and sizes are inches from the slide's top-left.
- Patches are atomic: every op is pre-validated (all errors reported at once, with the slide's real shape listing), then applied all-or-nothing.
**Full reference**: `python scripts/deck.py docs` — prints every op with semantics and recipes; needs no file. Read it before writing your first patch.
## The edit loop
```bash
# 1. READ — cheap orientation first, full JSON only when about to write a patch
python scripts/deck.py deck.pptx inspect --slide 3 --brief # one line per shape
python scripts/deck.py deck.pptx inspect --slide 3 # full JSON: text+formatting, image rIds, geometry, issues
python scripts/deck.py deck.pptx inspect --issues # only shapes with geometric problems
# 2. WRITE — one patch, many ops, atomic; chain repair + render of touched slides
python scripts/deck.py deck.pptx apply patch.json -o out.pptx --fix --render img/
# 3. VERIFY — look at what you changed
# (render names files slide-<0-based-index>.jpg to match inspect)
python scripts/deck.py deck.pptx diff out.pptx # structural changelog, no rendering
```
`--fix` runs deterministic geometry repair on the slides you touched (grows/shrinks/nudges overflowing text; never auto-moves pictures). Its *residue* report lists what still needs judgment, with a suggested op. Always look at the rendered image before declaring success.
## Patch ops at a glance
```json
{"ops": [
{"op":"set-text", "slide":3, "shape":"s12", "text":["Title", "Subtitle"]},
{"op":"swap-image", "slide":3, "shape":"s9", "image":"/abs/new.png"},
{"op":"swap-image", "media":"image13.png", "image":"/abs/logo.png"},
{"op":"replace-text","scope":"deck", "from":"Old Name", "to":"New Name"},
{"op":"replace-color","scope":"deck", "from":"E8A33D", "to":"F8DE6E"},
{"op":"set-theme", "colors":{"accent1":"BB7B19"}, "fonts":{"major":"Georgia"}},
{"op":"set-props", "title":"Q3 Review", "author":"Acme"},
{"op":"set-slide", "slide":3, "hidden":true, "background":"0F5258"},
{"op":"set-slide", "slide":3, "transition":{"type":"fade","speed":"med"}},
{"op":"set-notes", "slide":3, "notes":"speaker notes"},
{"op":"move", "slide":3, "shape":"s12", "to":[1.0,2.5]},
{"op":"resize", "slide":3, "shape":"s12", "size":[4.0,1.5]},
{"op":"set-style", "slide":3, "shape":"s12", "font_size":18, "fill":"0B3D3A", "rotation":6, "anchor":"MIDDLE"},
{"op":"delete", "slide":3, "shape":"s12"},
{"op":"duplicate", "slide":3, "shape":"s12", "offset":[0,1.2], "text":["Fourth pillar"]},
{"op":"copy-shape", "from_slide":8, "shape":"s12", "slide":3, "at":[1.0,2.0]},
{"op":"reorder", "slide":3, "shape":"s12", "z":"back"},
{"op":"add-shape", "slide":3, "kind":"textbox", "at":[1,2], "size":[4,1.5], "text":["…"], "name":"card"},
{"op":"add-picture", "slide":3, "image":"/abs/img.png", "at":[1,2], "width":4},
{"op":"add-table", "slide":3, "at":[1,2], "size":[8,3], "rows":[["A","B"],["1","2"]]},
{"op":"add-slide", "layout":"Blank", "at":5},
{"op":"add-row", "slide":3, "shape":"s12", "cells":["a","b"]},
{"op":"delete-col", "slide":3, "shape":"s12", "col":1}
]}
```
Key semantics (details in `docs`):
- **Vertical centering is `anchor`, not `alignment`**: `alignment` (set-text) is HORIZONTAL; to center text in a box taller than the text, set `"anchor":"MIDDLE"` (TOP|MIDDLE|BOTTOM) on set-style or set-text. `inspect` reports it when it's not the default top. html2patch sets it automatically from the slide's CSS (flex/grid centering).
- **set-text inherits formatting**: new paragraph *i* inherits ALL formatting of old paragraph *i* — pass plain strings for routine replacement. Pass objects to override (`{"text":"Big","font_size":28}`), or `"runs"` for mixed in-paragraph formatting. Table cells: add `"cell":[row,col]`.
- **Prefer duplicate/copy-shape over add-shape** when a styled donor exists — new shapes start from PowerPoint defaults, not the deck's design language.
- **add-shape `"name"`** lets later ops in the same patch target the shape it creates.
- **Transitions only on request**: never add slide transitions unless the user explicitly asks — applied uninvited or inconsistently they're annoying, not polish. When asked, default to one subtle type (fade) across the whole deck.
- **Alignment is linted, advisory**: `inspect`/`apply` report a `misaligned` issue when a shape's load-bearing edge sits a hair (0.03"–0.15") off a gridline ≥3 other shapes share — the near-miss that reads as sloppy. Act on it (move/resize the edge to the named coordinate) or ignore it: intentional asymmetry is real, so it never fails the build, and apply reports only a *newly introduced* near-miss.
- Keep new text comparable in length to the old, or let `fix` repair the overflow.
## Creating slides from HTML (html2patch)
When a slide should be DESIGNED from scratch — free-form layout, no styled
donor to duplicate — write it as HTML/CSS and compile it into a patch. The
browser is used as a measuring engine; the output is ordinary deck.py ops.
**Before writing any slide HTML, read [designing-slides.md](designing-slides.md)**
— how to design for this medium: subject-derived palettes, refusing the
default AI-deck looks, the token plan, projection type sizes, and what
survives compilation. The pipeline below is mechanical; that file is taste.
```bash
python scripts/html2patch.py slide.html --deck deck.pptx --layout Blank -o patch.json
python scripts/deck.py deck.pptx apply patch.json -o out.pptx --render img/
```
- The `<body>` is the slide at 96px/inch: 16:9 → `width:1280px; height:720px`.
Content overflowing the body is a compile error.
- Text lives in `<p>`, `<h1>`–`<h6>`, `<ul>`/`<ol>` (numbered), `<table>`, or
any element with inline-only content (figcaption, blockquote…). Divs with
background / border / border-radius / linear-gradient become styled rects;
`<img>` becomes a picture (local files only; `object-fit: cover` becomes a
real crop); `<table>` becomes a real table with per-cell fills and measured
column widths.
- Inline `<b>`/`<i>`/`<u>`/`<span style>` become formatted runs; `<a href>`
becomes a real hyperlink. CSS padding
maps to text insets; `text-transform`, `transform:rotate`, and vertical
`writing-mode` are honored. Use installed fonts (Arial, Georgia, …).
- By default each HTML file appends a new slide (`add-slide`, `--layout` picks
the template layout); `--slide N` targets an existing slide instead. Multiple
files compile into ONE atomic patch.
- Apply compiled patches WITHOUT `--fix` first — geometry is browser-measured,
so look at the render before repairing; run `fix` only if the render shows
a real problem.
- NEVER dismiss an overflow or `covered_by` flag on display-size text without
zooming that exact shape: `render --slide N --crop l,t,w,h --scale 2`.
Serif faces wrap differently in PowerPoint than in the browser — a clipped
last line is invisible at thumbnail size, especially when it falls behind
a picture.
- Extra dependenc提问、分享技巧或报告无法使用。按“有帮助”数量排序。