Claude Code Vision Skill
A Claude Code skill that adds multimodal vision to text-only model backends, letting the agent analyze pasted screenshots, UI states and charts through a configurable external vision model.
Claude Code Vision Skill gives Claude Code multimodal vision when its underlying model has none of its own, aimed at people running text-only models such as DeepSeek as their Claude Code backend. Pasting a screenshot with a keyboard shortcut sends it to a vision model for analysis without any extra command.
It supports Doubao, Qwen-VL, a DeepSeek vision variant, OpenAI's GPT-4o and Anthropic's Claude out of the box, plus any custom OpenAI- or Anthropic-compatible endpoint configured through per-provider environment variables, reaching self-hosted backends such as vLLM, Ollama, LiteLLM, OpenRouter or Azure OpenAI. Paired with a separate browser-automation tool, it can screenshot a rendered page, analyze layout and styling issues, and feed a review report back into the coding session.
It is written in Python, requires Python 3.10 or newer, and installs with pip.