add-image-vision

To add image vision to WhatsApp agent workflows, you merge the image-vision skill into your NanoClaw workspace, rebuild the container, and send a WhatsApp image to verify the agent can describe its contents.

Multimodal image processing for Claude agents works by downloading WhatsApp image attachments, resizing them with sharp, and packaging them as base64 content blocks that Claude can inspect alongside text inputs.

Yes, you can use sharp to preprocess WhatsApp images for multimodal interpretation by resizing and converting image attachments into base64 content blocks before sending them to Claude for visual inspection.

The best way to handle screenshot inputs in WhatsApp group chats is to use a skill that downloads, resizes, and packages image attachments as base64 multimodal content blocks so agents can reliably inspect and respond to visual data.

Your WhatsApp agent ignores image attachments because it lacks multimodal image processing capabilities. Adding a vision skill that converts images into base64 content blocks enables the agent to interpret and respond to visual inputs.

Yes, you need to rebuild the container after adding image vision to NanoClaw so that the sharp-based preprocessing, base64 multimodal content blocks, and channel integration are properly loaded for image-aware agent replies.