Skill UI

This skill autonomously optimizes artifacts like code or models using Hypothesis Tree Refinement. It runs iterative experiments with dev/test evaluators to prevent overfitting, coordinating subagents in isolated worktrees. Ideal for improving training recipes, agents, or pipelines without manual supervision.

Coordinate multiple agents as a team using work items with owners, agent Kanban state, branch/worktree isolation, control pane visibility, and explicit merge gates. Use for parallel agent fan-out and merging agent output into one product.