Mahal Systems WorkArena++
Fits browser agents doing compositional work in ITSM and HR portals. Borrow its way of building hard tasks by composing atomic ones and its ground-truth trace generator for training data.
What it asks the agent to do
- chain several ServiceNow actions into one workflow (e.g. find a record then update it)
- reason over list data and arithmetic before acting
- retrieve context from a knowledge base and apply it
- plan a multi-step task from a one-line request
What to borrow from it
case templatebaselinetaxonomy
Caveats
Same ServiceNow instance and gated-data requirement as WorkArena; a single vendor platform.
Details
| Workloads | it ops, workflow automation, web browsing, planning |
|---|---|
| Interaction | single turn |
| Environment | browser |
| End state checked | Yes |
| Tools | browser actions via BrowserGym |
| Grading | programmatic check |
| Size | 682 |
| Data | synthetic |
| Personal data | none |
| License | Apache-2.0 |
| Harness | own repo |
| Status | active |