Mahal Systems WebGames
Fits teams that want fast, hermetic checks of raw browser control. Borrow the password-on-success grading, which needs no judge or state inspection.
What it asks the agent to do
- solve a small self-contained browser challenge (drag, scroll, timing, canvas)
- handle tricky inputs, pop-ups and adversarial page tricks
- read visual puzzles on the page
What to borrow from it
grading designcase template
Caveats
Toy pages, not real sites; paper says 50+ challenges while the repo has 150; license not stated; best agent 43.1 percent versus 95.7 for humans.
Details
| Workloads | web browsing, games |
|---|---|
| Interaction | single turn |
| Environment | browser |
| End state checked | No |
| Tools | client side web pages, no network |
| Grading | exact match |
| Size | 150 |
| Data | synthetic |
| Personal data | none |
| License | unknown |
| Harness | own repo |
| Status | active |