Features
Linked datasets and scorers
A skill’s or an agent’s Evals tab now opens with its datasets and scorers: those linked to it, and those its past experiments ran against. Editors choose Add, then New dataset… or Link existing…, to add a dataset. Add scorer links an existing scorer or starts a new one.Improvements
- Overview: A skill’s Activity tab is now an Activity card on its Overview, listing its runs, saves, releases, and scans together.
- Source: A skill’s releases now show in a Versions list on Source, which replaces the Releases tab and marks the Live version.