pptx

---
name: pptx
description: Use when a .pptx file is involved in any way — creating decks, reading or extracting slide content, editing existing presentations, or working with templates, notes, and comments.
source: https://github.com/anthropics/skills/tree/main/skills/pptx
generated: 2026-07-02T18:43:03.361Z
category: tool
audience: general
---

## When to use

- Creating a slide deck, pitch deck, or presentation from scratch or a template
- Reading or extracting text from a .pptx, even to reuse the content elsewhere
- Editing, combining, or splitting existing presentations
- Working with templates, layouts, speaker notes, or comments in a deck

## Key concepts

### markitdown for reading

Text and content extraction runs through `python -m markitdown presentation.pptx`, which is also the tool used in content QA to catch missing content, typos, and leftover placeholders.

### Unpack / edit / pack workflow

Editing a template means analyzing it with thumbnail.py, then unpacking to raw XML, manipulating slides, editing content, cleaning, and repacking (detailed in editing.md).

### pptxgenjs for scratch decks

When no template or reference presentation exists, decks are built from scratch with the pptxgenjs Node library (see pptxgenjs.md).

### Design guidance

The skill pushes bold content-informed color palettes, a repeated visual motif, color dominance (60-70% one color), and a visual element on every slide — never plain title-plus-bullets.

### Mandatory QA bug hunt

QA is treated as an adversarial bug hunt with both content checks (markitdown, placeholder grep) and visual checks; success isn't declared until at least one fix-and-verify cycle completes.

### Subagent visual inspection

Slides are rendered to images via LibreOffice + pdftoppm and inspected by a fresh subagent, because the author's eyes see what they expect rather than what's there.

## API reference

```
npx skills add anthropics/skills --skill pptx
```

Install the PowerPoint (.pptx) skill.

```
npx skills add anthropics/skills --skill pptx
```

```
python -m markitdown presentation.pptx
```

Extract slide text for reading and content QA.

```
python -m markitdown output.pptx | grep -iE "xxxx|lorem|ipsum"
```

```
soffice.py --convert-to pdf + pdftoppm
```

Render slides to images for visual QA inspection.

```
python scripts/office/soffice.py --headless --convert-to pdf output.pptx
pdftoppm -jpeg -r 150 output.pdf slide
```

## Gotchas

- NEVER put accent lines under titles — they're a hallmark of AI-generated slides; use whitespace or background color instead
- Don't default to blue or Arial; pick a palette and font pairing informed by the specific topic
- Don't create text-only slides — every slide needs an image, chart, icon, or shape
- Your first render is almost never correct; do not declare success without a fix-and-verify cycle
- Use subagents for visual QA even for 2-3 slides — you'll see what you expect, not what's there
- After using a template, grep the output for leftover placeholder text (xxxx, lorem, ipsum) and fix before shipping

---
Generated by SkillMake from https://github.com/anthropics/skills/tree/main/skills/pptx on 2026-07-02T18:43:03.361Z.
Verify against source before relying on details.

File: ~/.claude/skills/pptx/SKILL.md