fix(ui-automation): Preserve runtime snapshot visibility contracts

Keep non-streaming UI action tests anchored to explicit success text
for every action case instead of accepting any non-empty output.

Avoid re-adding swipe actions to containers already marked offscreen by
viewport visibility filtering, and cover that ordering with a regression
test.

Co-Authored-By: Codex <noreply@openai.com>

Author: cameroncooke

src/mcp/tools/ui-automation/__tests__/non_streaming_progress.test.ts

@@ -85,6 +85,7 @@ describe('ui automation non-streaming tools', () => {
             axeHelpers,
           );
         },
+        expectedText: 'Long press on elementRef e1 for 1500ms simulated successfully.',
       },
       {
         name: 'swipe',
@@ -97,6 +98,7 @@ describe('ui automation non-streaming tools', () => {
             axeHelpers,
           );
         },
+        expectedText: 'Swipe up within elementRef e1 simulated successfully.',
       },
       {
         name: 'tap',
@@ -109,6 +111,7 @@ describe('ui automation non-streaming tools', () => {
             axeHelpers,
           );
         },
+        expectedText: 'Tap on elementRef e1 simulated successfully.',
       },
       {
         name: 'touch',
@@ -121,6 +124,7 @@ describe('ui automation non-streaming tools', () => {
             axeHelpers,
           );
         },
+        expectedText: 'Touch event (touch down) on elementRef e1 executed successfully.',
       },
       {
         name: 'type_text',
@@ -141,11 +145,7 @@ describe('ui automation non-streaming tools', () => {
       const { result } = await runToolLogic(testCase.run);
       expect(result.events, `${testCase.name} should not emit progress events`).toEqual([]);
       expect(result.isError()).toBe(false);
-      if (testCase.expectedText) {
-        expect(result.text()).toContain(testCase.expectedText);
-      } else {
-        expect(result.text().trim().length).toBeGreaterThan(0);
-      }
+      expect(result.text()).toContain(testCase.expectedText);
     }
   });
 

src/mcp/tools/ui-automation/__tests__/runtime-snapshot.test.ts

@@ -316,6 +316,50 @@ describe('runtime snapshot normalization', () => {
     );
   });
 
+  it('does not re-add swipeWithin to offscreen containers', () => {
+    const root = createNode({
+      type: 'Application',
+      role: 'AXApplication',
+      frame: { x: 0, y: 0, width: 390, height: 844 },
+      children: [
+        createNode({
+          type: 'Other',
+          role: 'AXGroup',
+          AXLabel: 'Offscreen panel',
+          frame: { x: 0, y: 900, width: 300, height: 200 },
+          children: [
+            createNode({
+              type: 'StaticText',
+              role: 'AXStaticText',
+              AXLabel: 'Overflowing child',
+              frame: { x: 10, y: 1160, width: 100, height: 20 },
+            }),
+          ],
+        }),
+      ],
+    });
+
+    const snapshot = createRuntimeSnapshotRecord({
+      simulatorId,
+      uiHierarchy: [root],
+      nowMs: 1_000,
+    });
+
+    expect(snapshot.payload.elements[1]).toEqual(
+      expect.objectContaining({
+        role: 'other',
+        label: 'Offscreen panel',
+        state: expect.objectContaining({ visible: false }),
+        actions: [],
+      }),
+    );
+    expect(snapshot.payload.actions).not.toContainEqual({
+      action: 'swipeWithin',
+      elementRef: 'e2',
+      label: 'Offscreen panel',
+    });
+  });
+
   it('removes point-based actions from clipped elements with offscreen activation points', () => {
     const root = createNode({
       type: 'Application',

src/mcp/tools/ui-automation/shared/runtime-snapshot.ts

@@ -407,6 +407,7 @@ function inferScrollableContainers(elements: RuntimeSnapshotElementRecord[]): vo
     const { publicElement, metadata } = element;
     if (
       !isContainerRole(publicElement.role) ||
+      publicElement.state?.visible === false ||
       !isVisible(publicElement.frame) ||
       !isLargeEnoughInferredScrollContainer(publicElement.role, publicElement.frame)
     ) {