Installing Anthropic Document Processing Skills in OpenClaw

The Anthropic skills repository provides high-quality document processing skills for PDF, PPTX, XLSX, and DOCX. These are knowledge-based skills (SKILL.md guides + helper scripts) that teach AI agents best practices for document handling. They do not require an Anthropic API key — they use standard open-source libraries.

What You Get

Skill	Capabilities
PDF	Extract text/tables (pdfplumber), merge/split/rotate (pypdf), create PDFs (reportlab), OCR scanned docs (tesseract), fill forms
PPTX	Read/extract text (markitdown), create slides (pptxgenjs), edit XML directly, convert to PDF/images (LibreOffice)
XLSX	Create/edit spreadsheets (openpyxl), data analysis (pandas), formula recalculation (LibreOffice)
DOCX	Create documents (docx npm package), read with pandoc, edit XML directly, handle tracked changes

Step 1: Install System Dependencies

# System tools
sudo apt install -y poppler-utils qpdf tesseract-ocr libreoffice pandoc imagemagick

# Python libraries
pip install pypdf pdfplumber reportlab pytesseract pdf2image openpyxl pandas "markitdown[pptx]" Pillow

# Node.js packages
npm install -g pptxgenjs docx

Dependency Matrix

Dependency	Used By	Purpose
`poppler-utils`	PDF, PPTX	pdftotext, pdfimages, pdftoppm
`qpdf`	PDF	PDF linearization, repair
`tesseract-ocr`	PDF	OCR for scanned documents
`libreoffice`	PPTX, XLSX, DOCX	Format conversion, formula recalc, accept tracked changes
`pandoc`	DOCX	Read content, extract tracked changes
`imagemagick`	PDF	Image processing
`pypdf`	PDF	Merge, split, rotate, encrypt
`pdfplumber`	PDF	Text and table extraction
`reportlab`	PDF	Create new PDFs
`openpyxl`	XLSX	Read/write Excel with formatting
`pandas`	XLSX	Data analysis
`pptxgenjs` (npm)	PPTX	Create presentations from scratch
`docx` (npm)	DOCX	Create Word documents from scratch

Step 2: Clone and Copy Skills

cd ~/.openclaw/workspace

# Clone the repo
git clone https://github.com/anthropics/skills.git anthropic-skills

# Create skill directories and copy
mkdir -p skills/pdf skills/pptx skills/xlsx skills/docx
cp -r anthropic-skills/skills/pdf/* skills/pdf/
cp -r anthropic-skills/skills/pptx/* skills/pptx/
cp -r anthropic-skills/skills/xlsx/* skills/xlsx/
cp -r anthropic-skills/skills/docx/* skills/docx/

# Clean up — remove the cloned repo
rm -rf anthropic-skills

Each skill directory should contain at minimum a SKILL.md and a scripts/ folder.

Step 3: Verify

openclaw skills list

You should see all four skills with status ✓ ready:

│ ✓ ready │ 📦 pdf  │ ...  │ openclaw-workspace │
│ ✓ ready │ 📦 pptx │ ...  │ openclaw-workspace │
│ ✓ ready │ 📦 xlsx │ ...  │ openclaw-workspace │
│ ✓ ready │ 📦 docx │ ...  │ openclaw-workspace │

Step 4: Restart OpenClaw

openclaw gateway restart

How It Works

These skills are not standalone tools — they are agent knowledge files. When your OpenClaw agent encounters a document task, it:

Reads the relevant SKILL.md for best practices and tool selection
Executes Python/Node.js commands via exec using the installed libraries
Follows the quality assurance steps defined in the skill

For example, when asked to “extract tables from a PDF”:

Agent reads: skills/pdf/SKILL.md
Agent runs:  python3 -c "import pdfplumber; ..."
Agent returns: extracted table data

Updating Skills

To pull the latest from Anthropic:

cd ~/.openclaw/workspace
git clone https://github.com/anthropics/skills.git anthropic-skills
cp -r anthropic-skills/skills/pdf/* skills/pdf/
cp -r anthropic-skills/skills/pptx/* skills/pptx/
cp -r anthropic-skills/skills/xlsx/* skills/xlsx/
cp -r anthropic-skills/skills/docx/* skills/docx/
rm -rf anthropic-skills

Notes

No Anthropic API key needed — these skills use only open-source tools
Linux fully compatible — all dependencies are standard packages
LibreOffice is the heaviest dependency — required for PPTX/XLSX/DOCX format conversion and formula recalculation; can be skipped if you only need PDF
The shared scripts/office/soffice.py wrapper handles LibreOffice headless mode automatically