/ui-review

You are running the visual UI review loop on this target: $ARGUMENTS

Follow the ui-visual-review skill for the procedure and the ui-design-standards skill for the rubric. Do not skip the render step — you must look at a real image, never review from source alone.

Steps:

1. Locate the widget/dialog/panel in the codebase. If the user gave a screenshot path instead, view that image directly and skip to step 3.
2. Render it to a PNG with the offscreen harness (scripts/snap.py in the ui-visual-review skill). Apply the project's real theme. Render at the real on-screen size, and in the states that bite: empty / typical / overflowing, and — for dockable panels — at a narrow width (~280–320 px) to catch overflow and disappearing buttons.
3. View every PNG you produced.
4. Critique against ui-design-standards. For each violation, name the rule, the exact widget, and the concrete fix (with numbers).
5. Fix the source. Prefer the smallest change that satisfies the rule.
6. Re-render and re-view to confirm the fix landed and introduced no regression. Repeat until the screen passes the rubric.
7. Report a short before/after summary with the rule IDs you resolved.

If the app is already instrumented with qt-mcp and running, you may use the live qt_* tools instead of the offscreen harness for steps 2–6 (see the qt-mcp.md reference). Otherwise, the offscreen harness needs no running app.