[mgoldsborough]

A kit overlay's scrim can only cover the app's frame, never the host chrome around it. At 0.4 (dialog) and 0.32 (drawer), the frame's edge read as a break in the page while an overlay was open. Both now use one constant, SCRIM_BACKGROUND (rgba(0, 0, 0, 0.2)) in internal/top-layer.ts; the panel's border and shadow carry the separation.

• Test: the dialog and drawer stylesheets carry the same scrim colour.
• CHANGELOG: Unreleased → Changed.
• npm run ci green (657 tests); the vendored IIFE is unchanged (the UI kit is not in that bundle).

[mgoldsborough]

QA Review: PR #134

Scope: 5 files, +47/−4 · worktree: supplied review worktree; creation and dependency installation skipped as requested · reviewed: 2888606 · CI: green
Risk: LOW
Why this exists: Reduce the visual discontinuity between an embedded app's dimmed frame and surrounding host chrome, using one scrim colour for both overlays.

Critical (must fix)

None.

Fix in-PR (apply mechanically — no adjudication essay expected)

None.

Suggestions (optional)

None.

What Looks Good

• Both components consume the same internal constant; public APIs, modal interactions, and stacking behavior are unchanged. This is a non-breaking visual change.
• The added test checks the injected stylesheets against the shared value and each other, following the repository's DOM-testing guidance.
• All supplied required and additional CI checks pass. No suite was rerun locally, and no files were changed.

Body claims

• ✅ Both scrims now use rgba(0, 0, 0, 0.2), replacing 0.4 and 0.32 respectively.
• ✅ A stylesheet regression test and an Unreleased → Changed changelog entry were added.
• ✅ The vendored IIFE is unchanged in the diff; the passing Build check includes its freshness comparison.
• The supplied checks confirm passing test jobs on Node 22 and 24, but do not expose the claimed total of 657 tests or the author's local command history. Visual improvement was not independently assessed in a browser.

machine record

{
  "kind": "review",
  "pr": 134,
  "round": 1,
  "reviewed": "2888606952c12b1cd2f32bbe436c3fcdd31283de",
  "delta_base": null,
  "ci": "green",
  "risk": "LOW",
  "stop_gate": "n/a",
  "verdict": "No findings; eligible for qa-reviewed.",
  "label": "eligible-on-green",
  "findings": []
}

[mgoldsborough]

QA Review: PR #134

Scope: Delta: 11 files, +65/−25; branch against origin/main: 5 files, +47/−4 · supplied review worktree; creation and dependency installation skipped · reviewed: 9264277 · CI: green
Round: 2 (delta base: 2888606)
Stop gate: SHIP THE CORE
Risk: LOW

Critical (must fix)

None.

Fix in-PR (apply mechanically — no adjudication essay expected)

None.

Suggestions (optional)

None.

What Looks Good

• The delta merges origin/main's gutter configuration and version update. The shared lighter scrim change remains intact.
• Gutter defaults, public exports, stylesheet assertions, documentation, and JavaScript/Python client version pins agree.
• Round 1 had no findings requiring fix verification. The full branch severity-floor scan found no data-loss, security, or authorization concerns.
• All prefetched required and additional CI checks pass. No local suite was rerun and no worktree files were changed.

Body claims

Not re-audited in round 2, per the delta-review scope.

machine record

{
  "kind": "review",
  "pr": 134,
  "round": 2,
  "reviewed": "92642775c116e34153ac8cdb50316a7a06f193cd",
  "delta_base": "2888606952c12b1cd2f32bbe436c3fcdd31283de",
  "ci": "green",
  "risk": "LOW",
  "stop_gate": "ship-the-core",
  "verdict": "SHIP THE CORE — no findings.",
  "label": "eligible-on-green",
  "findings": []
}