markdown-converter

name	markdown-converter
description	Convert documents and files to Markdown using markitdown. Use when converting PDF, Word (.docx), PowerPoint (.pptx), Excel (.xlsx, .xls), HTML, CSV, JSON, XML, images (with EXIF/OCR), audio (with transcription), ZIP archives, YouTube URLs, or EPubs to Markdown format for LLM processing or text analysis.

Markdown Converter

Convert files to Markdown using uvx markitdown — no installation required.

Basic Usage

# Convert to stdout
uvx markitdown input.pdf

# Save to file
uvx markitdown input.pdf -o output.md
uvx markitdown input.docx > output.md

# From stdin
cat input.pdf | uvx markitdown

Supported Formats

• Documents: PDF, Word (.docx), PowerPoint (.pptx), Excel (.xlsx, .xls)
• Web/Data: HTML, CSV, JSON, XML
• Media: Images (EXIF + OCR), Audio (EXIF + transcription)
• Other: ZIP (iterates contents), YouTube URLs, EPub

Options

-o OUTPUT      # Output file
-x EXTENSION   # Hint file extension (for stdin)
-m MIME_TYPE   # Hint MIME type
-c CHARSET     # Hint charset (e.g., UTF-8)
-d             # Use Azure Document Intelligence
-e ENDPOINT    # Document Intelligence endpoint
--use-plugins  # Enable 3rd-party plugins
--list-plugins # Show installed plugins

Examples

# Convert Word document
uvx markitdown report.docx -o report.md

# Convert Excel spreadsheet
uvx markitdown data.xlsx > data.md

# Convert PowerPoint presentation
uvx markitdown slides.pptx -o slides.md

# Convert with file type hint (for stdin)
cat document | uvx markitdown -x .pdf > output.md

# Use Azure Document Intelligence for better PDF extraction
uvx markitdown scan.pdf -d -e "https://your-resource.cognitiveservices.azure.com/"

Notes

• Output preserves document structure: headings, tables, lists, links
• First run caches dependencies; subsequent runs are faster
• For complex PDFs with poor extraction, use -d with Azure Document Intelligence

子代理执行策略

• 默认执行模式：优先使用子代理独立执行；主会话负责接单、补齐必要上下文、控制范围，并在最后整合结果回报给用户。
• 适合交给子代理的任务：预计超过 30 秒的处理、多步骤流水线、多文件/多产物生成、批处理、审计、转写、总结、离线分析。
• 主会话保留职责：只在必要时追问关键缺口、确认边界、挑选最终结果，并把输出改写成面向用户的完成答复。
• 不建议默认交给子代理的情况：一次性极短任务、需要高频来回确认的对话、强依赖当前聊天即时上下文的动作。
• 回报节点：默认只保留“已受理 / 已开始 / 已完成 / 已失败”四类关键节点，避免刷屏。