| name | html-slides-to-pdf |
| description | Converts single-file HTML slide decks into a multi-page PDF for WeChat, email, or print. Runs a headless browser, one PDF page per `.slide`, merges with pypdf. Use when the user wants PDF from HTML slides, export slides to PDF, share slides on WeChat, or print presentations. |
HTML Slides → PDF (Agent Skill)
This repository is agent-first: this file is SKILL.md at the repo root. Install by cloning the repo into your agent’s skills folder (see README.md → Install the agent skill), then install the Python + Playwright dependencies so the agent can run html_to_pdf.py.
Supported setups (same skill file, different install paths):
| Agent | Typical skill location | Notes |
|---|
| Cursor Agent | ~/.cursor/skills/html-slides-to-pdf/SKILL.md | One-liner: git clone https://github.com/yuanmu97/html-slides-to-pdf.git ~/.cursor/skills/html-slides-to-pdf (see README). |
| Claude Code | ~/.claude/skills/html-slides-to-pdf/SKILL.md | One-liner: git clone https://github.com/yuanmu97/html-slides-to-pdf.git ~/.claude/skills/html-slides-to-pdf |
The canonical skill file in git is SKILL.md at the repository root.
Prerequisites (one-time, machine)
Use one Python for both pip install and running the script:
cd /path/to/html-slides-to-pdf
python -m pip install -r requirements.txt
python -m playwright install chromium
What the agent should run
Replace /path/to/html-slides-to-pdf with the actual clone path:
python /path/to/html-slides-to-pdf/html_to_pdf.py "/path/to/deck.html" -o "/path/to/deck.pdf"
If the package was installed with pip install -e ., the html-slides-to-pdf command is equivalent to python .../html_to_pdf.py.
Defaults: viewport 1920×1080, 800 ms delay per slide. Override with --width, --height, --delay-ms. Debug: --keep-temp.
Rendering details (screen media, vh pinning, flex-safe slide toggling) are in README.md under How it works.
Behavior summary
- Slide detection: If
.reveal .slides exists, uses reveal.js Reveal.getSlides() when loaded, otherwise top-level section children (static export). Otherwise uses .slide elements, excluding the reveal root (div.reveal that contains .slides) so reveal’s slide class does not collapse the deck to one page.
- Dot-slide decks: Toggles
.active on the current panel so patterns like .slide { display:none } + .slide.active { display:flex } (JS-driven decks) export every page.
- Screen media; pins document/slide size to
--width×--height px; reveal layouts get transform reset so slides fill the viewport.
print-color-adjust: exact and copying computed background onto html/body helps dark themes and gradients appear in PDF.- Hides
.nav-dots, .progress-bar, .keyboard-hint, reveal controls/progress, body > .controls / body > .page-indicator, .export-btn, etc.
If something goes wrong
No module named 'playwright' — Same python for pip install and html_to_pdf.py.- No slides — Need reveal
.slides > section, or .slide panels (not only the reveal wrapper).
- Clipped / fonts / timing — Adjust
--width, --height, --delay-ms.
- Layout vs browser — Match
--width / --height to the viewport you design in (default 1920×1080).
Repo layout (for the agent)
html_to_pdf.py — main scriptrequirements.txt — dependenciesexamples/slides/mobile-ai-runtime.html — example deck (16 slides)
