Turn an outline into a deck people can actually present
Skills used in this guide (3)
只要涉及 .pptx 或 .potx 文件（作为输入、输出或两者）时使用。包括：创建幻灯片、路演稿或演示；读取、解析或提取任意 .pptx/.potx 文本（即便提取内容将用于别处，如邮件或摘要）；编辑、修改或更新现有演示；合并或拆分幻灯片文件；使用模板（.potx）、版式、演讲者备注或批注。当用户提及 deck、slides、presentation 或引用 .pptx/.potx 文件名时触发，无论其后续用途。若需打开、创建或改动 .pptx/.potx，请使用本技能。
npx skills add anthropics/skills -g为产物应用主题样式的工具集。产物可以是幻灯片、文档、报表、HTML 落地页等。内置 10 套预设主题（含配色与字体），可应用到任何已创建的产物，或即时生成新主题。
npx skills add anthropics/skills -g将 Anthropic 官方品牌配色与字体应用于任何可能受益于 Anthropic 视觉风格的产物。当涉及品牌配色、样式规范、视觉排版或公司设计标准时启用。
npx skills add anthropics/skills -g
The usual failure is not "AI cannot make slides". It is slides where every page is a paragraph: dense text, no focus, and the presenter ends up reading aloud. This workflow splits what to say from how it looks, with a manual trim in between.
Who this is for
Good fit
- You already have content (a plan, a weekly report, a course outline, research findings) and need it on screen
- 8 to 30 page decks for reviews, readouts or talks
- You are willing to delete words: one conclusion per slide
Bad fit
- Brand-grade keynotes with complex animation, masters and linked charts
- Content you have not figured out yet
- Pixel-perfect corporate templates (hand the template file over first)
What to prepare
- A Markdown outline. One conclusion per line, not full sentences.
- The goal: who is listening, how many minutes, what decision you want.
- Optional: the company
.pptxtemplate, placed in the same folder.
Steps
1. Install the skill
git clone --depth 1 --filter=blob:none --sparse https://github.com/anthropics/skills
cd skills && git sparse-checkout set skills/pptxCopy skills/pptx/ into your client skills directory (~/.claude/skills/, ~/.codex/skills/, ~/.cursor/skills/, ~/.trae/skills/, ~/.codebuddy/skills/, ~/.lingma/skills/).
2. Compress the outline before generating
Rewrite this outline, do not generate a file yet:
1. One conclusion per content slide; cut anything over 20 words.
2. Max 3 bullets per slide; split if more.
3. Mark slides that need a chart at the end of the line: [chart: line / compare / flow].
4. List what you removed so I can confirm.3. Generate the .pptx
Ask for 16:9, titles above 28pt, body above 18pt, max 3 bullets, placeholder boxes for charts, and a final "Next steps" slide with owner and date. Then open the file and check for text overflow and font fallback.
4. Apply one theme
Use a theme skill for a single accent colour, or paste your brand hex values into the prompt.
5. Add charts and speaker notes
Ask for the numbers as a table you can paste, rather than asking the model to draw SVG. Ask for 40-word speaker notes that say the one thing the slide is for.
Traps
| Trap | Symptom | Fix |
|---|---|---|
| Walls of text | Looks like pasted Word | Hard-code "max 3 bullets, body >= 18pt" |
| CJK font fallback | Slides open in the wrong font | Reset the font manually after generation |
| Page count explodes | 8 bullets become 30 slides | Compress the outline first |
| Invented numbers | Charts look credible but have no source | Require [TODO] where data is missing |
| No animation | Slides are static | Readout decks rarely need animation; add manually if required |
FAQ
Can I hand it a Word doc? Yes, but extract an outline first. Direct conversion gives you paragraphs on slides.
Does the .pptx open in Keynote or WPS? Yes. Keynote may shift spacing, so export a PDF for important talks.
How many words per slide? Title under 15 characters, body under 3 lines of 20 characters. More means two slides.
Can I reuse our template? Put template.pptx in the working folder and say "use as master". If master names differ, save a blank template first.
Batch production? Polish one deck by hand, then have AI swap content using that structure.
References
Related Scenes
Write PRDs and product specs with AI
The pain of PRD isn't "wrong format" — it's "goal and acceptance unclear, dev guesses". The point: give background, target user, problem; let AI output "feature list + acceptance + edges", not decide "why build" for you.
Write weekly reports and standups with AI
The pain of weekly reports isn't "can't write" — it's "turning a log into outcomes". The point: give AI the raw material (what you did, what's stuck), have it close into "progress / risk / next", not invent content for you.
Translate and localize docs with AI
Machine translation's scare is "fluent but distorted" — wrong terms, wrong tone, stiff local phrasing. The point: give a glossary and target-reader tone, have it "translate + keep proper nouns", then you verify key sentences — not trust the whole blindly.
Handle your inbox and support messages with AI
The red line for email automation is "draft yes, auto-send careful". "Triage + draft + you hit send" is safest; letting AI blast-send means you own the mistake, not it.