openclaw-magika#

AI-powered file content type detection using Google Magika.

Quick Use#

bash

# Single file — label output
magika -l /path/to/file

# MIME type output
magika -i /path/to/file

# Recursive directory scan
magika -r ./workspace/

# JSON output (for programmatic use)
magika --json /path/to/file

# Batch multiple files
magika -r --json ./uploads/

Common Tasks#

TaskCommandIdentify single filemagika -l <file>Get MIME typemagika -i <file>Recursive scanmagika -r <dir>Batch JSONmagika --jsonl <files...>With confidence scoremagika -s -l <file>

Output Labels#

Magika identifies 200+ content types including: markdown, json, python, javascript, pdf, zip, png, jpeg, html, xml, sql, csv, png, gif, mp3, mp4, exe, elf, wasm, and many more.

High-confidence predictions are marked with [HIGH]. Low-confidence returns generic labels like text or unknown binary data.

Install#

bash

# pipx (recommended)
pipx install magika

# pip
pip install magika

# brew
brew install magika

# curl script
curl -LsSf https://securityresearch.google/magika/install.sh | sh

Notes#

• Fast: ~5ms per file after model loads
• Accurate: 99% average precision on 200+ file types
• Local: All inference runs locally, no network needed
• Size agnostic: Only reads a subset of file content regardless of file size