super-ux

What problem does it solve?

super-ux eliminates the manual, time-consuming work of checking whether a live UI matches a reference design by repeatedly running a builder-reviewer cycle until screenshots are visually clean.

Core Features & Use Cases

• Autonomous design-fidelity iteration: Executes N rounds of Builder↔Reviewer ping-pong, where the Builder applies targeted code fixes and the Reviewer measures pixel diffs against reference screenshots.
• Screenshot + pixel-diff driven findings: Produces structured YAML findings and updates loop state, halting on clean status, halt conditions, or iteration limits.
• Git worktree isolation for safe iterations: Applies fixes on separate worktrees per role per iteration, merges results into an aggregate branch, and keeps the Reviewer read-only on source code.

Use Case example: You maintain a design system and want overnight visual QA across all pages listed in docs/design-loop/targets.md, stopping automatically when the UI matches the committed reference images.

Quick Start

Ask your agent to run /super-ux with a target loop size, for example: "Run /super-ux 8 to iterate design fixes and pixel-diff the app against the reference screenshots until the UI is clean."