markitdown · jimmc414 — Claude Code skill: convert files to Markdown for LLMs
markitdown — a Claude Code skill that turns PDFs, Office documents, images, audio, and web pages into Markdown ready for language models. Handles 20+ formats while keeping headings, lists, tables, and links intact.
no license 1 installs ★ 607 updated 2026-10-09
npx skills add jimmc414/Kosmos --skill markitdown -a claude-codeWhat markitdown does
Built on the Python utility MarkItDown: feed it a file or URL, get back clean Markdown with no temporary files left behind.
Converts PDF, DOCX, PPTX, XLSX, HTML, EPUB, CSV, JSON, and XML. Images go through OCR with EXIF metadata pulled out; audio gets transcribed via the speech_recognition package. A YouTube link yields its transcript. ZIP archives are unpacked whole — every file inside is converted and concatenated into one result.
For tricky PDFs there's Azure Document Intelligence integration, which extracts tables and layout more accurately. You can also wire in GPT-4o to describe images inside presentations. The extension plugin system ships disabled by default for security and turns on with enable_plugins=True.
What it helps with
- extract text from PDF or Office files
- run OCR on images along with EXIF data
- transcribe audio to text
- pull YouTube transcripts from a URL
- batch-convert documents for RAG or embedding pipelines
Example prompts for the agent
- Convert
manual.pdfto Markdown and save it asmanual.md - Process every file in
documents/and write Markdown versions tomarkdown/ - Unpack
archive.zipand convert all PDFs inside to Markdown
How to install markitdown
1Skills CLI (Vercel)
npx skills add jimmc414/Kosmos --skill markitdown -a claude-code2Manually: put the files in place
~/.claude/skills/markitdown/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/markitdown/
1Skills CLI (Vercel)
npx skills add jimmc414/Kosmos --skill markitdown -a codex2Manually: put the files in place
.agents/skills/markitdown/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add jimmc414/Kosmos --skill markitdown -a cursor2Manually: put the files in place
.cursor/skills/markitdown/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install jimmc414/Kosmos kosmos-claude-scientific-skills/scientific-skills/markitdown2Manually: put the files in place
.github/skills/markitdown/SKILL.md1Skills CLI (Vercel)
npx skills add jimmc414/Kosmos --skill markitdown -a gemini-cli2Platform CLI
gemini skills install https://github.com/jimmc414/Kosmos --path kosmos-claude-scientific-skills/scientific-skills/markitdownManually: put the files in place
.devin/skills/markitdown/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add jimmc414/Kosmos --skill markitdown -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about markitdown
How do I install markitdown in Claude Code?
Run npx skills add jimmc414/Kosmos --skill markitdown -a claude-code. The skill lands in ~/.claude/skills/markitdown/SKILL.md.
What does markitdown do?
Converts PDFs, Office files, images, audio, HTML, and structured data into Markdown tuned for LLM consumption. Headings, lists, tables, and hyperlinks survive the conversion.
What do I need for audio transcription?
The speech_recognition package. For better table extraction from PDFs you need an Azure Document Intelligence endpoint and key; for GPT-4o image descriptions, an OpenAI client.
Similar skills
All skills for Agent setup: memory, context, orchestration →markitdown by other authors comes first.