How to Use Claude Code Skills for Document Processing: A Complete Guide

Claude Code skills are reusable bundles that extend Claude Code's capabilities, enabling it to read, edit, and generate Office documents, PDFs, spreadsheets, and presentations through a local MCP server.

Claude Code skills transform how developers handle document workflows by bringing powerful automation directly into the terminal. This guide covers everything you need to know about using Claude Code skills for document processing, from installation to advanced orchestration techniques.

What Are Claude Code Skills?

Claude Code skills are self-contained packages that define a contract for specialized tasks. Each skill contains a skill.json manifest, runtime code, and descriptions of inputs and outputs. When installed, skills reside in ~/.claude/skills/ and expose their functionality through a local MCP (Model Context Protocol) server.

The MCP server provides a language-agnostic JSON-over-HTTP API that runs skill code in an isolated sandbox. This architecture ensures sensitive documents never leave your machine while enabling seamless integration with the Claude Code CLI.

Official Document Processing Skills

Anthropic maintains four official document processing skills listed in the VoltAgent/awesome-agent-skills repository under the Official Claude Skills section:

Skill	Extension	Capabilities
anthropics/docx	.docx	Create, edit, and analyze Word documents
anthropics/pdf	.pdf	Extract text, generate PDFs, handle forms
anthropics/pptx	.pptx	Create, edit, and analyze presentations
anthropics/xlsx	.xlsx	Read, edit, and analyze spreadsheets

These skills share a consistent interaction pattern: the CLI sends a JSON payload to the MCP endpoint, and the skill returns either modified files or extracted data.

Installing Document Processing Skills

All skills install through the unified claude skill install command. The skill name uses the format owner/skill-name from the official catalog.

# Install the Word document skill

claude skill install anthropics/docx

# Install multiple document skills at once

claude skill install anthropics/pdf anthropics/xlsx anthropics/pptx

Installation copies the skill bundle to ~/.claude/skills/anthropics/skill-name/. You can verify installation by checking the directory or listing available skills.

Core Document Processing Operations

Extracting Text from PDFs

The PDF skill's extract-text action converts PDFs to plain text, preserving structure where possible.

claude skill run anthropics/pdf extract-text \
    --file ./contracts/agreement.pdf \
    --output ./extracted/agreement.txt

The --output parameter specifies where the extracted text file should be written. The skill processes the PDF in the sandbox and returns only the text content.

Generating Word Documents from Prompts

The generate action creates structured documents from natural language descriptions.

claude skill run anthropics/docx generate \
    --prompt "Create a quarterly report for Q1 2026 with executive summary, revenue table, and two placeholder charts" \
    --output ./reports/q1-2026.docx

The skill interprets the prompt, applies appropriate formatting, and generates a complete Office Open XML document. You can refine results by adding more specific instructions about styling, sections, or content structure.

Editing Excel Spreadsheets

The XLSX skill supports targeted cell updates, formula application, and range operations.

claude skill run anthropics/xlsx edit \
    --file ./data/metrics.xlsx \
    --range "B2:C10" \
    --formula "SUM(A2:A10)" \
    --output ./data/metrics-updated.xlsx

Key parameters:

• --range: Target cell range in A1 notation
• --formula: Formula to apply (optional; omit for value-based updates)
• --output: Path for the modified workbook

Creating PowerPoint Presentations

Generate slide decks from structured outlines with theme support.

claude skill run anthropics/pptx generate \
    --outline "1. Project Overview\n2. Current Challenges\n3. Proposed Solution\n4. Implementation Timeline\n5. Budget and Resources\n6. Next Steps" \
    --theme "Modern" \
    --output ./presentations/project-brief.pptx

The skill parses the numbered outline into slide titles, applies the selected theme, and adds appropriate placeholders for content expansion.

Advanced Orchestration Patterns

Combining Multiple Skills in Shell Scripts

Document workflows often require chaining skill operations. This example extracts PDF content, generates a summary, and creates a formatted report:

#!/usr/bin/env bash
set -e

# Configuration

pdf_source="./docs/specifications.pdf"
temp_text="./tmp/specs.txt"
final_report="./reports/specs-summary.docx"

# Ensure temp directory exists

mkdir -p "$(dirname "$temp_text")"

# Step 1: Extract text from PDF

claude skill run anthropics/pdf extract-text \
    --file "$pdf_source" \
    --output "$temp_text"

# Step 2: Generate summary using Claude's LLM capabilities

summary=$(claude ask "Summarize the technical specifications in $temp_text in under 200 words, focusing on system requirements and performance benchmarks.")

# Step 3: Create formatted Word document

claude skill run anthropics/docx generate \
    --prompt "Create a professional summary document titled 'Technical Specifications Review'. Include: $summary. Add a header with date and footer with page numbers." \
    --output "$final_report"

echo "Report generated: $final_report"

This pattern demonstrates how Claude Code skills for document processing integrate with the broader Claude ecosystem, combining deterministic skill operations with LLM reasoning.

Parameterizing Document Generation

For production workflows, externalize configuration to enable template reuse:

#!/usr/bin/env bash

# Source configuration

source ./config/report-config.sh

# Validate required variables

[[ -z "$REPORT_TITLE" ]] && { echo "Error: REPORT_TITLE not set"; exit 1; }
[[ -z "$DATA_SOURCE" ]] && { echo "Error: DATA_SOURCE not set"; exit 1; }

# Generate with validated parameters

claude skill run anthropics/docx generate \
    --prompt "Generate $REPORT_TYPE titled '$REPORT_TITLE' using data from $DATA_SOURCE. Style: $REPORT_STYLE. Sections: $REPORT_SECTIONS" \
    --output "$OUTPUT_PATH"

Externalizing configuration separates document structure from business logic, enabling consistent formatting across multiple outputs.

Key Files and References

Understanding the repository structure helps navigate the Claude Code skills for document processing ecosystem:

File	Purpose	Location
README.md	Official catalog of all document-processing skills with installation commands	Repository root
README.md (lines 96-100)	Specific entries for anthropics/docx, anthropics/pdf, anthropics/pptx, anthropics/xlsx	Lines 96-100
~/.claude/skills/	Local installation directory for all skills (runtime path, not in repository)	Local filesystem
skill.json (per-skill)	Manifest defining actions, input schemas, and MCP endpoints	Inside each installed skill bundle

The VoltAgent/awesome-agent-skills repository serves as the definitive reference for discovering and installing official document processing capabilities.

Summary

Claude Code skills for document processing provide a secure, local-first approach to automating Office and PDF workflows:

• Install skills using claude skill install anthropics/<skill-name> for docx, pdf, pptx, or xlsx handling
• Run operations via claude skill run with action-specific parameters for generation, extraction, or editing
• Process locally through the MCP server sandbox—documents never leave your machine
• Orchestrate workflows by chaining skills in shell scripts, combining deterministic operations with LLM reasoning
• Reference the catalog at VoltAgent/awesome-agent-skills for official skill documentation

Frequently Asked Questions

How do I install multiple Claude Code document processing skills at once?

You can install multiple skills in a single command by listing them as separate arguments: claude skill install anthropics/docx anthropics/pdf anthropics/xlsx anthropics/pptx. Each skill downloads to ~/.claude/skills/anthropics/<skill-name> with its complete skill.json manifest and runtime code.

What file formats does the Claude Code PDF skill support?

The anthropics/pdf skill handles standard PDF documents for text extraction and form processing. It outputs plain text files that preserve structural formatting where possible. For PDF generation, the skill creates new documents from structured content rather than modifying existing files in place.

Can I edit existing Excel files with Claude Code skills?

Yes. The anthropics/xlsx skill supports targeted edits through the edit action with parameters including --file for the source workbook, --range for cell selection in A1 notation, and --formula for applying calculations. The skill always writes to a new output file rather than overwriting the original.

Where are Claude Code skills stored after installation?

Installed skills reside in ~/.claude/skills/ with each skill organized under its owner namespace (e.g., ~/.claude/skills/anthropics/docx). This location is checked at runtime when resolving skill names. The repository cache at .claude/skills/ in your current project directory can also hold project-specific skill installations.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}