Hwp Pipeline

hwp-pipeline

A [Claude Code](https://docs.anthropic.com/en/docs/claude-code) skill for reading, converting, and editing Korean Hangul documents (`.hwp` / `.hwpx`) — without Hancom Office.

Why

Korean government forms, academic templates, and institutional documents are distributed as `.hwp` files. Editing them typically requires Hancom Office (Windows-only subscription). This skill provides a complete pipeline for working with HWP/HWPX on any OS:

.hwp → [read] → JSON / Markdown / HTML
.hwp → [convert] → .hwpx → [edit] → filled form

Features

Capability	Backend	Requires
Read HWP (JSON, Markdown, HTML)	@ohah/hwpjs	Node.js 18+
Convert HWP to HWPX	hwp2hwpx	Java 11+, Maven
Edit HWPX (create, template fill, replace)	python-hwpx	Python 3.10+

Quick start

1. Install

git clone https://github.com/Yoojin-nam/hwp-pipeline.git
cp -r hwp-pipeline ~/.claude/skills/hwp-pipeline
cd ~/.claude/skills/hwp-pipeline
bash setup.sh

2. Use with Claude Code

The skill activates automatically when you mention HWP/HWPX files or Korean document editing. Examples:

• "Convert this HWP to Markdown"
• "Fill in the placeholders in this government form"
• "Extract text from this HWPX file"

3. Use scripts directly

**Extract text from HWPX:**

python3 scripts/text_extract.py document.hwpx --include-nested

**Replace placeholders:**

python3 scripts/zip_replace_all.py template.hwpx output.hwpx \
  --replace "{name}=John" "{date}=2026-01-01" --auto-fix-ns

**Convert HWP to HWPX:**

java -jar java/hwp2hwpx-fat.jar input.hwp output.hwpx

Project structure

hwp-pipeline/
├── SKILL.md              # Claude Code skill definition
├── setup.sh              # One-command dependency installer
├── java/
│   ├── Convert.java      # HWP→HWPX