Markitdown Converter

MarkItDown Converter Skill

Convert files (PDF, DOCX, XLSX, PPTX, HTML, CSV, images, audio) to Markdown using Microsoft's MarkItDown library.

Usage

Skill({ skill: 'markitdown-converter' });
// Then call the converter via Bash:
// bash: python .claude/tools/cli/markitdown-convert.py <input_file> [output_file]

Supported File Types

Installation

pip install 'markitdown[all]'

CLI Wrapper

The CLI wrapper is at .claude/tools/cli/markitdown-convert.py.

Convert a file to stdout

python .claude/tools/cli/markitdown-convert.py /path/to/file.pdf

Convert and save to output file

python .claude/tools/cli/markitdown-convert.py /path/to/file.pdf /path/to/output.md

Convert with explicit extension hint (for streams)

python .claude/tools/cli/markitdown-convert.py /path/to/file --ext .pdf

Exit Codes

JSON Output Format

{
  "success": true,
  "text_content": "# Document Title\n\nContent here...",
  "char_count": 1234,
  "source_file": "report.pdf",
  "output_file": null
}

Integration with Telegram File Drop

When a user sends a file in Telegram:

// 1. Download the file to tmp
const tmpPath = `.claude/context/tmp/telegram-upload-${userId}-${Date.now()}.${ext}`;

// 2. Run markitdown converter
const result = JSON.parse(
  execSync(`python .claude/tools/cli/markitdown-convert.py "${tmpPath}"`).toString()
);

// 3. Store as agent memory
if (result.success) {
  MemoryRecord({ type: 'discovery', text: result.text_content.slice(0, 2000), area: 'user-files' });
}

Error Handling

// Check if markitdown is installed
const { status } = spawnSync('python', ['.claude/tools/cli/markitdown-convert.py', '--help']);
if (status === 3) {
  // Not installed — guide user
  sendMessage(chatId, 'File conversion requires: pip install markitdown[all]');
}

When to Use

• User drops a file in Telegram → convert to markdown → store as memory
• Agent needs to process uploaded documents
• Convert research papers (PDF) to searchable markdown
• Extract data from spreadsheets (XLSX → markdown tables)