Axiom Vision Diag

This is the diagnostics companion to the axiom-vision skill family, specifically handling when Vision framework features break or behave unexpectedly. You'd reach for this when subject detection fails, pose landmarks go missing, text recognition returns garbage, or your DataScanner shows a blank screen. It covers the common failure modes like subjects at frame edges, confidence threshold tuning, coordinate conversion bugs between lower-left and top-left origins, and performance issues that freeze your UI. The skill assumes you already know which Vision API you're using and just need to fix why it's not working as expected.