Claude Qwen Vision

Handles images when the session model is blind to them (e.g. DeepSeek): the user's message contains an image — pasted into the terminal (the [Image #N] placeholder) or an image file path — convert it to text with qwen3.8-max via scripts/qwen_vision.py and answer from that text instead of trying to view the image directly. Triggers — 用户贴图/截图/paste an image, 「看下这张图」「读出图里的文字」「图片在 xxx.png」. If user_prompt_hook is installed, pasted images arrive already described — skip.

sunfmin 9498722 7 files · 37.1 KB Updated

File contents

sunfmin/claude-qwen-vision/tree/main/skills/claude-qwen-vision commit 949872233f

Frequently asked questions

npx skillmds@latest add sunfmin/claude-qwen-vision