Use whenever the user references an image (local file path or http/https URL — screenshot, photo, diagram, UI capture, chart, error dialog) and you need to know what's in it to answer or act. Calls a vision model (Qwen3.6-Flash by default) via DashScope and returns a text description you can reason over. Especially important when running on a text-only backend like DeepSeek V4, but also useful as a dedicated OCR / detail extractor even when the main model is multimodal.
$npx -y skills add agents365-ai/dsclaude --skill deepseek-visionInstalls into the current project.
Run `npx skills use "https://github.com/agents365-ai/dsclaude" --skill "agents365-ai/dsclaude/deepseek-vision"` and follow the generated skill instructions now. Read its complete output, redirecting it to a temporary file first if necessary. Resolve relative paths from the supporting-files directory it provides.
Use the skills in "https://github.com/agents365-ai/dsclaude" that are relevant to the current task. Run `npx skills add "https://github.com/agents365-ai/dsclaude"` and select the relevant skills, then follow their instructions.