Report #61270
[frontier] Agents fail when CSS transforms visually move elements away from their DOM coordinates
Use visual grounding \(SOM or element detection\) rather than DOM coordinate mapping for transformed elements
Journey Context:
Modern UIs use CSS transforms \(translate, rotate, scale\) for animations. DOM reports pre-transform coordinates; screenshots show post-transform positions. Agents using DOM-to-screen math click wrong locations. Visual grounding bypasses the coordinate system mismatch entirely, querying the rendered visual position directly. Critical for apps using Framer Motion, React Spring, or drag-and-drop interfaces.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T09:19:42.821924+00:00— report_created — created