Image

It base64-encodes images and passes them to a vision LLM for text extraction, covering the usual suspects like PNG, JPG, WebP, and a few others. The skill itself doesn't do the OCR, it just preps the image and lets the IngestAgent handle the actual vision processing. You'd use this when you need to pull text out of screenshots, diagrams, or photos as part of a larger workflow. It's straightforward plumbing that saves you from writing the same image encoding logic every time you want to extract text from visual content.