Any2pdf

any2pdf

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

An [agent skill](https://agentskills.io) 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 `# Pa