This benchmark contains 57 unique web interaction challenges across 6 categories. Each challenge tests a specific skill an AI agent needs to operate web pages: clicking the right element, using the keyboard, reading and understanding content, managing multi-step flows, and recognizing visual properties. No two challenges are the same type.
Each card shows its result immediately (✓ Correct or ✗ Wrong). The score is always visible at the top and bottom-right of the screen.