siliconflow-vision

What This Skill Does

siliconflow-vision is a powerful image analysis agent skill designed for OpenClaw. It acts as the "eyes" of your AI assistant, converting raw visual data into structured, objective, and detailed descriptive output. Unlike generic image upload features, this skill is engineered to trigger deep analysis, specifically designed to extract text, identify screen elements, define layout structures, and interpret artistic or informational styles. It bridges the gap between raw pixel data and high-level reasoning by providing the main AI model with factual, noise-free visual information before any synthesis or interpretation occurs. With native support for SiliconFlow (default), OpenAI, and Anthropic, it ensures high compatibility and flexibility in choosing the underlying vision model.

Installation

To integrate this skill into your OpenClaw ecosystem, execute the following command in your terminal:

clawhub install openclaw/skills/skills/lycohana/siliconflow-vision

After installation, ensure your API keys are configured correctly by updating the config/default.json file or setting the corresponding environment variables (SILICONFLOW_API_KEY, OPENAI_API_KEY, or ANTHROPIC_API_KEY) to enable the desired service provider.

Use Cases

• Developer Support: Analyze screenshots of error logs, stack traces, or UI layout bugs in desktop/mobile apps.
• Content Creation: Interpret complex memes or visual humor by letting the model extract the literal elements, while the main agent handles the cultural context.
• Office Productivity: Digitize data from charts, extract text from scanned documents, or summarize information from meeting slides.
• Data Analysis: Extract numerical data points from infographics or business reports for further processing.

Example Prompts

1. "[Image] Could you explain what is wrong with this Python code screenshot?"
2. "[Image] Extract the main data points from this chart and summarize the trend for my weekly report."
3. "[Image] Analyze this meme. What are the visual components and why does it have this specific satirical tone?"

Tips & Limitations

• Tip: Always specify the -m smart flag for high-precision tasks like reading technical documentation or complex infographics to leverage the Qwen-VL-72B model.
• Tip: For faster responses when dealing with simple screenshots, use the -s (short) mode to reduce output tokens.
• Limitation: The skill is designed for objective description; avoid asking it to perform high-level subjective analysis directly, as it performs best when the primary model handles the final synthesis and external research.