Agentic UI: Fix Linux/Windows window controls

[gavande1]

Related issues

• Related to STU-2030

How AI was used in this PR

AI helped locate the window-control code across the main process and both renderers, diagnose the height and color mismatch, and draft the fix. I reviewed the code and verified the behavior by hand in the running Linux app, across light and dark mode and both UI modes.

Proposed Changes

On Windows and Linux, the native window controls (minimize, maximize, close) in the agentic UI overflowed into the site preview and did not follow the theme, so in light mode they showed as an out-of-place dark block.

Now the preview toolbar reserves the exact area the OS reports for the controls, so they no longer spill into the site view, and the controls match the toolbar color in both light and dark mode. The color also stays in sync when you toggle the theme or switch between the classic and agentic UI. The classic UI keeps its existing dark controls, so it is unchanged.

Testing Instructions

1. Run on Linux or Windows
2. Open the agentic UI with a running site and its preview visible.
3. Confirm the window controls sit within the preview toolbar and do not spill into the site view.
4. Confirm the controls match the toolbar color in both light and dark mode, with a gap between the controls and the preview option icons.
5. Toggle the theme and switch between the classic and agentic UI, and confirm the controls recolor to match each time.
6. Confirm the classic UI controls are unchanged, and that macOS is unaffected.

Pre-merge Checklist

• Have you checked for TypeScript, React or other console errors?

[wpmobilebot]

📊 Performance Test Results

Comparing 4a77db8 vs trunk

app-size

Metric	trunk	4a77db8	Diff	Change
App Size (Mac)	1419.48 MB	1419.48 MB	+0.00 MB	⚪ 0.0%

site-editor

Metric	trunk	4a77db8	Diff	Change
load	1070 ms	1078 ms	+8 ms	⚪ 0.0%

site-startup

Metric	trunk	4a77db8	Diff	Change
siteCreation	6967 ms	7012 ms	+45 ms	⚪ 0.0%
siteStartup	2354 ms	2350 ms	4 ms	⚪ 0.0%

---

Results are median values from multiple test runs.

Legend: 🟢 Improvement (faster) | 🔴 Regression (slower) | ⚪ No change (<50ms diff)