Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
This skill provides a bridge for LLMs that cannot directly process visual data. By utilizing a helper script, it transforms images from local paths, URLs, or the system clipboard into text-based descriptions and analyses.
It is designed for developers who need to recognize content within images or analyze visual references while working in a text-only model environment.
WHO IT'S FOR
AI agent / automation builders
integrating image analysis into LLM workflows
Platform / DevEx teams
extending AI assistant capabilities with local tools
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Follow the documented setup, then try a first task.
Setup
Local clone with Claude Code configuration
Installs: Configures image recognition in an ordinary Claude Code project by copying and configuring repository components; the repository is cloned locally first. · Claude Code
Before you start
Git must be available to run the documented clone command.
Choose a supported vision service and model.
Obtain the selected service's API key.
For a non-Qwen service, obtain its OpenAI-compatible API base URL.
Node.js must be available to run vision.js.
Run from: The directory where the claude-vision-skill repository should be cloned and from which Claude Code can access its relative path.
In a Claude Code session, enter this exact request: 读一下 claude-vision-skill/README.md，帮我配置识图. This is an agent-session request, not a terminal command.
When Claude Code asks, provide the chosen service, API key, and—if using a non-Qwen service—the API base URL. The documented project setup places vision.js in the project root, substitutes the key and model in vision.js, changes BASE_URL when necessary, and merges CLAUDE.md into the project's CLAUDE.md.
Installs: Configures image recognition in an ordinary Claude Code project by asking Claude Code to follow the repository README directly. · Claude Code
Before you start
Claude Code must be able to access the GitHub repository; the README warns that some third-party models may not be able to access GitHub.
Choose a supported vision service and model.
Obtain the selected service's API key.
For a non-Qwen service, obtain its OpenAI-compatible API base URL.
Node.js must be available to run vision.js.
Run from: The Claude Code project to configure.
In a Claude Code session, enter this exact request: 按 https://github.com/asuojun/claude-vision-skill 的 README 帮我配置识图. This is an agent-session request, not a terminal command.
When Claude Code asks, provide the chosen service, API key, and—if using a non-Qwen service—the API base URL. The documented project setup places vision.js in the project root, substitutes the key and model in vision.js, changes BASE_URL when necessary, and merges CLAUDE.md into the project's CLAUDE.md.
Replace <absolute image path> with the image file's absolute path.
Ensure the .env file next to vision.js contains DASHSCOPE_API_KEY, VISION_MODEL, and DASHSCOPE_BASE_URL as appropriate; do not paste the API key into the prompt.
A Node.js runtime is needed to run the documented node command.
Make sure claude-vision-skill is installed in your agent before trying this task.
Use claude-vision-skill to give me a concise Chinese description of the main contents of the image at <absolute image path>. Mention any clearly visible text and important objects, but do not infer details that are not visible.
Claude
Codex
DeepSeek