Skills

Luobo PPT: Generate and Beautify Editable Native PPTX in Your Agent

For turning an article or topic into an editable native PPTX — or beautifying an existing .pptx — inside Agents that can execute commands, such as Claude Code, Codex, Gemini CLI, and Cursor. Before use, confirm local dependencies like Python 3.8+ and a Chinese font; the first run automatically installs missing Python packages.

Official source · luobo-ppt

Install

git clone https://github.com/zhouwei713/luobo-ppt.git ~/.claude/skills/luobo-ppt

Who It's For

  • People who need to quickly turn an article or a topic into an editable, further-modifiable PPTX file — especially users working in Agents that can execute commands, such as Claude Code, Codex, Gemini CLI, and Cursor.
  • It's also for those who already have a .pptx and want to beautify it as a whole: the repository clearly offers two routes — beautify an existing .pptx, or generate from an article/topic.
  • Prerequisites: Python 3.8+ locally, plus python-pptx, Pillow, numpy, lxml, and a Chinese font; LibreOffice is recommended for previews. Missing Python packages will be installed automatically by the skill on first use. The runtime environment is a local environment with a usable command line.
  • Not suitable or unknown cases: the repository does not state whether pure web-based Agents are supported, whether slides can be generated in English or other languages, or whether there are Office version compatibility limits — treat these as unknown and do not assume.

First Try in 3 Minutes

The recommended main path is the official Claude Code installation method:

  1. Confirm prerequisites: Python 3.8+ and a Chinese font are installed (LibreOffice is also recommended for previews).
  2. Installation location: clone it into the current Agent's skills directory. The official command for Claude Code:
git clone https://github.com/zhouwei713/luobo-ppt.git ~/.claude/skills/luobo-ppt
  1. Trigger the task: give instructions directly in natural language in a supported Agent (the repository states it works in Agents that can execute commands, such as Claude Code, Codex, Gemini CLI, and Cursor — just say it). For the first run, choose a low-risk, observable task: give it a short topic from your own article and ask it to generate a small outline-style PPT.
  2. Side effects and checkpoints: generation proceeds through the pipeline — plan.json, creating illustrations, assembling the native pptx, rendering previews, automatic checks, a five-dimension self-evaluation, and delivery. The first run will install missing Python packages (it downloads dependencies over the network) and writes the generated .pptx to the output location you specify. This is an operation that writes new files, so confirm the output directory first.
  3. Success signals: following the official pipeline, you should see an assembled native pptx, rendered previews, and the skill's five-dimension self-evaluation. Note: this is a document/presentation generation pipeline; it does not involve specialized review rubrics for novels or professional documents. The repository does not explain the specific meaning of its self-evaluation dimensions, which is unknown.

For Codex and other Agents supporting the open SKILL.md standard, the officially given secondary method is to clone it into ~/.agents/skills/luobo-ppt. Installation methods for other platforms are not stated in the evidence — do not infer them yourself.

Use Cases

Case 1: Article to PPT. Paste an article of a few hundred words to the Agent and say, "Use luobo-ppt to generate an editable PPT from this article." The skill runs the full pipeline: generating plan.json, creating illustrations, assembling the native pptx, rendering previews with automatic checks, and finally delivering the file and the five-dimension self-evaluation. What you get is not a collage of images but a natively editable pptx whose text you can keep editing in Office software.

Case 2: Beautify existing slides. Give the Agent the path to an old .pptx and say, "Please beautify this PPT." This takes the "beautify the original .pptx" route explicitly listed in the repository, and likewise goes through the pipeline from plan.json to assembly, preview, self-evaluation, and delivery. The evidence does not state whether beautification overwrites the original file — back up the original first and ask the Agent to explicitly output to a new file.

Usage Boundaries

  • Files and commands: the skill executes Python-related commands locally and writes the generated .pptx plus intermediate files (plan.json, deck.json, style.json, preview images, etc.). First use will install missing Python packages. Whether it reads or writes other files in your working directory is not stated in the evidence.
  • Network: whether installing dependencies and possibly generating illustrations accesses external services is not stated in the repository — unknown.
  • Checkpoints: before any action that could overwrite an existing .pptx, clarify the output path first; there is no official evidence that beautification is lossless.
  • Runtime and clients: a local Python 3.8+ environment; the Agent must support open standards like SKILL.md and be able to execute commands (the official examples are Claude Code, Codex, Gemini CLI, and Cursor, but the repository does not say it only works in these four Agents — whether other similar Agents work is unknown).
  • Officially stated limitations: a Chinese font is required; LibreOffice is recommended for previews (a suggestion, not a requirement).
  • Unverified key information: the specific five self-evaluation dimensions, the range of supported languages for generated slides, the overwrite policy for existing files, and availability on pure web-based Agents — none of these are covered in the repository documentation; rely on actual runs before use.