submit → screen → canonize → test → publish
Arena
Suggest a real task worth measuring. The submission is screened and rewritten into a fixed statement and acceptance criteria before any run.
Queue
The public queue is empty
A submitted task can remain in screening until a maintainer writes its canonical statement and acceptance criteria.
What screening rejects
- a duplicate of a task already in the queue
- hidden executables, download links, or destructive shell commands
- a task that requires a submitter's private account or private data
- a task whose acceptance criteria cannot be written before the run