v1.0.0

A fully-local image understanding skill for Claude Code, powered by macOS Vision Framework. Helps models without vision capabilities (e.g. DeepSeek v4 Flash) read text, extract tables, and describe image content.

Features

• Fully local — no network, images never leave your Mac
• OCR — Chinese & English text extraction in reading order
• Table detection — --layout reconstructs table structure with coordinates
• Content description — --describe turns textless images into structured text
• QR / barcode decoding — decoded content output directly

Requirements

• macOS 11.0+
• Xcode Command Line Tools

Install

git clone git@github.com:niyongsheng/free-vision-skill.git
mkdir -p ~/.claude/skills/free-vision-skill
cp -r SKILL.md scripts examples ~/.claude/skills/free-vision-skill/