any2pdf

Tell your AI assistant “转PDF” and get a publication-quality document. No config, no templates, no LaTeX.

An agent skill that gives AI coding assistants (Claude Code, Cursor, Copilot, Gemini CLI, etc.) the ability to convert Markdown into professionally typeset PDFs — with a single natural language request.

Why This Exists

Every existing Markdown-to-PDF tool falls into one of two traps:

1. Too simple — Pandoc/wkhtmltopdf produce passable output for English, but CJK mixed text gets butchered: wrong fonts, broken line wraps, “Claude Code” split across lines, 年月日 rendered as □□□.

2. Too complex — LaTeX produces beautiful output but requires a 4GB TeX distribution, arcane syntax, and breaks on every CJK edge case. No AI assistant can reliably drive it.

any2pdf is the sweet spot: one Python file, one dependency (reportlab), zero config — and it handles every CJK/Latin edge case because we hit them all building real 200-page Chinese technical reports.

What Makes It Different

For AI Agents — Zero-Friction Workflow

The skill teaches your AI assistant an interactive workflow, not just a CLI command:

You: "把这个报告转成PDF"

Agent: asks about design style, frontispiece, watermark, back cover
Agent: runs the conversion with all your choices

One sentence from you → a complete, branded document.

For CJK Users — Battle-Tested on Real Reports

Every fix in this codebase came from a real rendering bug in a real report:

• Mixed text wrapping: “Chaofan Shou” won’t split across lines; Chinese text breaks naturally at character boundaries
• Canvas CJK: Dates like “2026年4月1日” render correctly everywhere — cover, headers, footers
• Book-quality fonts: Palatino + 宋体 (macOS), Times + SimSun (Windows) — not Arial + fallback
• Rich Markdown blocks: Obsidian callouts, inline images, emoji, and LaTeX-style formulas render instead of disappearing
• Merged heading recovery: Input like # Part## Chapter on one line? Auto-split before parsing

For Everyone — Cross-Platform, Zero Config

Fonts are auto-discovered from system paths. Missing fonts? You get a helpful error with the exact install command for your OS.

On Ubuntu/Debian, the broadest setup is:

sudo apt install fonts-dejavu-core fonts-liberation fonts-freefont-ttf fonts-noto fonts-noto-cjk fonts-noto-color-emoji

Install

npx skills add skill-publisher/any2pdf -g -y

Works with 25+ AI agents: Claude Code, Cursor, GitHub Copilot, Gemini CLI, Codex, Cline, Warp, and more.

Design Styles

10 built-in themes — from classic LaTeX-inspired to modern minimalist. Click to preview.

What You Get

• Cover page with title, subtitle, author, version, stats lines
• Clickable table of contents with PDF bookmark sidebar
• Frontispiece — full-page image after cover (AI-generated or local)
• Running headers — stable report/document title without page-lagged chapter labels
• Running footers — author/brand, page number, date
• Watermark — faint diagonal text on every content page
• Back cover — banner image or text branding (QR codes, business cards)
• 10 design themes — from warm academic to ink wash minimalist
• Markdown images — local and remote ![alt](src) images with graceful fallback text
• Obsidian callouts — > [!NOTE], warnings, tips, quotes, and related callout types
• LaTeX-style formulas — inline $...$ and display $$...$$ / \[...\]; install matplotlib for rendered math images
• Emoji fallback — Twemoji image rendering when online, with local emoji-font fallback when available
• YAML frontmatter — set title, theme, watermark, cover, TOC, and other options inside the Markdown file

Direct CLI Usage

You can keep document options in frontmatter so the source stays reproducible:

---
title: My Report
author: Author Name
theme: warm-academic
watermark: DRAFT
---

pip install reportlab
# Optional, for rendered formula images instead of styled fallback text:
pip install matplotlib

python lov-any2pdf/scripts/md2pdf.py \
  --input report.md \
  --output report.pdf \
  --title "My Report" \
  --author "Author Name" \
  --theme warm-academic \
  --watermark "DRAFT" \
  --toc true

See SKILL.md for the full 20+ argument reference.

License

MIT