Pandoc PDF Generation

Overview

Generate professional PDF documents from Markdown using Pandoc with the XeLaTeX engine. This skill covers automatic section numbering, table of contents, bibliography management, LaTeX customization, and common troubleshooting patterns learned through production use.

When to Use This Skill

Use this skill when:

• Converting Markdown to PDF with professional formatting requirements
• Needing automatic section numbering and table of contents
• Managing citations and bibliographies without manual duplication
• Controlling table formatting and page breaks in LaTeX output
• Building automated PDF generation workflows

Quick Start: Universal Build Script

Single Source of Truth Pattern

This skill provides production-proven assets in ${CLAUDE_PLUGIN_ROOT}/skills/pandoc-pdf-generation/assets/:

• table-spacing-template.tex – Production-tuned LaTeX preamble (booktabs, colortbl, ToC fixes)
• build-pdf.sh – Universal auto-detecting build script

From Any Project

/usr/bin/env bash << 'DETECT_EOF'
# Create symlink once per project (git-friendly)
ln -s ${CLAUDE_PLUGIN_ROOT}/skills/pandoc-pdf-generation/assets/build-pdf.sh build-pdf.sh

# Auto-detect single .md file in directory (landscape default)
./build-pdf.sh

# Portrait mode
./build-pdf.sh --portrait document.md

# Monospace font for ASCII diagrams
./build-pdf.sh --monospace diagrams.md

# Explicit input/output
./build-pdf.sh input.md output.pdf
DETECT_EOF

Options:

Features:

• ✅ Auto-detects input file (if single .md exists)
• ✅ Auto-detects bibliography (references.bib) and CSL files
• ✅ Always uses production-proven LaTeX preamble from skill
• ✅ Pre-flight checks (pandoc, xelatex, files exist)
• ✅ Post-build validation (file size, page count)
• ✅ Code blocks stay on same page (no splitting across pages)
• ✅ Lua filter to hide <details> blocks from PDF output

Landscape PDF (Quick Command)

For landscape PDFs with blue hyperlinks (no build-pdf.sh dependency):

pandoc file.md -o file.pdf \
  --pdf-engine=xelatex \
  -V geometry:a4paper,landscape \
  -V geometry:margin=1in \
  -V fontsize=11pt \
  -V mainfont="DejaVu Sans" \
  -V colorlinks=true \
  -V linkcolor=blue \
  -V urlcolor=blue \
  --toc --toc-depth=2 \
  --number-sections

Use landscape for: Wide data tables, comparison matrices, technical docs with code blocks.

Manual Command (With LaTeX Preamble)

/usr/bin/env bash << 'SKILL_SCRIPT_EOF'
pandoc document.md \
  -o document.pdf \
  --pdf-engine=xelatex \
  --toc \
  --toc-depth=3 \
  --number-sections \
  -V geometry:margin=1in \
  -V mainfont="DejaVu Sans" \
  -H ${CLAUDE_PLUGIN_ROOT}/skills/pandoc-pdf-generation/assets/table-spacing-template.tex
SKILL_SCRIPT_EOF

ASCII Diagrams: Always Use graph-easy

CRITICAL: Never manually type ASCII diagrams. Always use the itp:graph-easy skill.

Manual ASCII art causes alignment issues in PDFs. The graph-easy skill ensures:

• Proper boxart character alignment
• Consistent spacing
• Reproducible output

# Invoke the skill for general diagrams
Skill(itp:graph-easy)

# For ADR architecture diagrams
Skill(itp:adr-graph-easy-architect)

Also important: Keep annotations OUTSIDE code blocks. Don’t add inline comments like # contains: file1, file2 inside diagram code blocks – they break alignment.

Hiding Content for PDF Output

Use --hide-details to remove <details> blocks from PDF output. This is useful when:

• graph-easy source blocks: Keep source in markdown for diagram regeneration, but hide from printed PDFs
• Technical implementation notes: Show in web/markdown view, hide from printed handouts
• Collapsible sections: HTML <details> tags don’t render as collapsible in PDF

Usage:

./build-pdf.sh --hide-details document.md

Markdown pattern:

## My Section

```diagram
┌─────┐     ┌─────┐
│ Box │ ──> │ Box │
└─────┘     └─────┘
```

[Box] -> [Box]

With --hide-details, the entire <details> block is stripped from PDF output while remaining visible in markdown/HTML.

Verification Checklist

Before considering a PDF “done”, verify:

Pre-Generation:

• No manual section numbering in markdown (use --number-sections)
• All ASCII diagrams generated via itp:graph-easy skill
• Annotations are outside code blocks, not inside

Post-Generation:

• Open PDF and visually inspect each page
• Verify diagrams don’t break across pages
• Check section numbering is correct (no “1. 1. Title” duplication)
• Confirm bullet lists render as bullets, not inline dashes

Pre-Print:

• Get user approval before printing
• Confirm orientation preference (landscape/portrait)
• Confirm duplex preference (one-sided/two-sided)

Printing Workflow

Always let the user review the PDF before printing.

Open for review:

open output.pdf

Print one-sided (simplex):

lpr -P "PRINTER_NAME" -o Duplex=None output.pdf

Print two-sided (duplex):

lpr -P "PRINTER_NAME" -o Duplex=DuplexNoTumble output.pdf  # Long-edge binding
lpr -P "PRINTER_NAME" -o Duplex=DuplexTumble output.pdf    # Short-edge binding

Find printer name:

lpstat -p -d

Never print without user approval – this wastes paper if issues exist.

Reference Documentation

For detailed information, see:

• Core Development Principles – START HERE – Universal principles learned from production failures
• Markdown for PDF – Markdown structure patterns for clean landscape PDFs
• YAML Front Matter Structure – YAML metadata patterns
• LaTeX Customization – Preamble and table formatting
• Bibliography & Citations – BibTeX and CSL styles
• Document Patterns – Document type templates
• Troubleshooting – Common issues and fixes

Troubleshooting