SkillsBench challenge; Terminal-Bench 2 out of the arena; gates check SkillsBench; practice off the board

Per-task credit metadata and content hashes; rank on the mean of verified runs

Whoami honours Bearer HF tokens and reports is_editor

Dogfood visibility: organizer pipeline runs and base-model row on the dashboard, board posts on run events, BenchFlow-native task packages accepted, challenge runs serve on four GPUs

Document verified submitted-environment HF training and publication

Run pinned submitted environments on HF with evidence collection and reviewed result publication

Use Agent Collabs dashboard with environment submissions and configurable experiments