AgentSkillsCN

markitdown

将 PDF、Office 文件、图像、音频以及网页内容转换为干净的 Markdown，以供 LLM 处理。适用于需要阅读、提取文本、转录，或以适合上下文的格式处理文件时使用。

SKILL.md
--- frontmatter
name: markitdown
description: >-
  Converts PDF, Office, Images, Audio, and Web content to clean Markdown for LLM processing.
  Use when needing to read, extract text, transcribe, or process files into context-friendly format.
license: Apache-2.0
metadata:
  version: "1.0.0"
  trit: 0
  author: agent-builder

MarkItDown

Convert various file formats into Markdown optimized for LLM processing.

📚 References

ReferencePurpose
references/document_conversion.mdPDF, DOCX, PPTX, XLSX conversion
references/media_processing.mdImage OCR and Audio transcription
references/web_content.mdHTML, YouTube, EPUB extraction
references/structured_data.mdCSV, JSON, XML to Markdown tables
references/advanced_integrations.mdAzure Doc Intelligence & LLM Image descriptions

🚀 Quick Start

Basic Conversion:

python
from markitdown import MarkItDown
md = MarkItDown()
result = md.convert("document.pdf")
print(result.text_content)

Command Line:

bash
markitdown document.pdf -o output.md

YouTube Transcript:

python
result = md.convert("https://youtube.com/watch?v=VIDEO_ID")

📦 Installation

Full installation:

bash
uv pip install 'markitdown[all]'

Modular installation:

bash
uv pip install 'markitdown[pdf]'    # PDF
uv pip install 'markitdown[docx]'   # Word
uv pip install 'markitdown[audio]'  # Audio