Document Processing Skill

A Claude Code skill and MCP server for converting scattered meeting materials into structured Word documents.

For a detailed explanation of the design and development of this code, see the related article on Data Science Collective.

Overview

This project provides an automated pipeline for transforming various inputs (audio recordings, handwritten notes, diagrams, digital notes, and supplementary documents) into a single, well-formatted Microsoft Word deliverable.

Project Structure

document-processing-skill/
├── documenting-meetings/          # Claude skill for meeting documentation
│   ├── SKILL.md                   # Main skill specification and workflow
│   ├── EVALUATION.md              # Test evaluation prompts and criteria
│   └── reference/
│       ├── INPUT_FORMATS.md       # Detailed input file handling guide
│       └── OUTPUT_SECTIONS.md     # Output section specification
├── transcription-MCP/             # MCP server for audio/video transcription
│   ├── server.py                  # FastMCP server implementation
│   └── .env                       # Environment configuration (requires setup)
└── sample_data/                   # Example data for testing
    ├── input_documents/           # Meeting materials (audio, images, docs, notes)
    ├── templates/                 # Blank template documents
    └── sample_documents/          # Sample output documents for formatting reference

Components

1. Claude Skill: documenting-meetings

A Claude Code skill that orchestrates the entire meeting documentation workflow.

Trigger Keywords: meeting notes, meeting summary, meeting minutes, meeting documentation, action items from meeting

Supported Input Formats:

Workflow:

1. Collect context (meeting title, output preferences, focus areas)
2. Validate input folder structure
3. Inventory input files
4. Transcribe audio/video recordings via MCP
5. Interpret images (handwritten notes, diagrams)
6. Read digital notes
7. Check for templates and sample documents
8. Generate Word deliverable
9. Save to input folder

Default Output Structure:

• Meeting Summary (date, attendees, duration)
• Executive Summary
• Decisions Made
• Action Items (table with owner, due date, priority)
• Open Questions
• Follow-up Message (email template)

Sections are omitted if no relevant information exists.

2. MCP Server: transcription-MCP

A FastMCP server providing audio/video transcription capabilities using the GAIK transcriber library and OpenAI API.

Tool Exposed:

transcribe_audio(file_path: str, enhanced: bool = False) -> str

Parameters:

• file_path (required): Full path to audio/video file
• enhanced (optional): Return enhanced transcript if True

Returns: Raw transcription text preserving original flow and structure.

Setup

Prerequisites

• Python 3.8+
• OpenAI API key
• Claude Code with MCP support
• GAIK transcriber library

MCP Server Configuration

1. Navigate to the transcription-MCP folder:

   cd transcription-MCP
   

2. Create/update the .env file with your OpenAI API key:

   OPENAI_API_KEY=your_openai_api_key
   OPENAI_API_TYPE=openai
   

3. Install dependencies:

   pip install mcp python-dotenv gaik
   

4. Register the MCP server in your Claude Code configuration.

Skill Installation

1. Copy the documenting-meetings folder to your Claude Cod

... View full README on GitHub