document-SKILLs

Document Processing Skills

English | 中文

A collection of Claude Code / Codex skills for document manipulation — PDF extraction/forms, Excel analysis/formulas, Word, and PowerPoint.

Quick start

Claude Code, as a plugin

Two commands inside Claude Code, no cloning:

/plugin marketplace add appautomaton/document-SKILLs
/plugin install document-skills@appautomaton-skills

Skills are then invoked as document-skills:pdf, document-skills:docx, and so on. The system tools under Dependencies are still required; the docx and pptx skills install their Node packages the first time they need them.

Any agent, via symlinks

[!TIP] Clone this repo anywhere you keep your skills, then symlink each skill into your agent's skills directory.

git clone https://github.com/appautomaton/document-SKILLs.git ~/skills/documents
cd ~/skills/documents

# Claude Code
for s in docx pdf pptx xlsx; do
  ln -s "$(pwd)/$s" ~/.claude/skills/$s
done

# Codex
for s in docx pdf pptx xlsx; do
  ln -s "$(pwd)/$s" ~/.codex/skills/$s
done

Skills

Every workflow checks its own output before handing it back: workbooks are recalculated and scanned for errors, Word and PowerPoint files are validated against the OOXML schemas, and documents, PDFs and decks are rendered so the agent can look at them.

Dependencies

[!NOTE] Python scripts use uv + PEP 723 inline metadata — no virtual environments or pip install needed. Just uv run.

System tools:

# macOS: one command via the repo Brewfile.
# Installs uv and node only if missing, optional tools are commented in the file.
brew bundle

# Linux
sudo apt-get install -y pandoc poppler-utils libreoffice
sudo apt-get install -y tesseract-ocr   # optional, OCR of scanned PDFs

Node.js packages (docx and pptx skills only):

cd docx && npm install
cd ../pptx && npm install

Development

tests/ runs every script through uv run, the way an agent does. See AGENTS.md.

License

MIT — see LICENSE.

🤖 Checkout linux.do for more fun stuff about AI!