test(mothership): a desktop tool's executor resolves only once the tool has settled

Author: waleedlatif1

apps/sim/lib/mothership/tools/client/browser-tool-execution.test.ts

@@ -332,6 +332,35 @@ describe('executeBrowserToolOnClient', () => {
     })
   })
 
+  it('resolves only once the browser action has settled and its result is handed to delivery', async () => {
+    let finishAction: (result: unknown) => void = () => {}
+    mockExecuteBrowserTool.mockReturnValue(
+      new Promise((resolve) => {
+        finishAction = resolve
+      })
+    )
+    const toolCallId = nextToolCallId()
+    let resolved = false
+    const running = executeBrowserToolOnClient(toolCallId, 'browser_snapshot', {}, CHAT_SCOPE).then(
+      () => {
+        resolved = true
+      }
+    )
+
+    await vi.waitFor(() => expect(mockExecuteBrowserTool).toHaveBeenCalled())
+    await sleep(10)
+    expect(resolved).toBe(false)
+    expect(mockReportCompletion).not.toHaveBeenCalled()
+
+    finishAction({ text: 'page content' })
+    await running
+    // Delivery itself is owned by the retained-completion scheduler, which outlives the turn's
+    // Stop and the page; what the caller waits for is the action, with its report under way.
+    expect(mockReportCompletion).toHaveBeenCalledWith(toolCallId, 'success', expect.any(String), {
+      text: 'page content',
+    })
+  })
+
   it('lets a running invocation own the genuine result when the same call is re-delivered', async () => {
     let finishExecution: (result: { text: string }) => void = () => {}
     mockExecuteBrowserTool.mockImplementation(

apps/sim/lib/mothership/tools/client/local-filesystem.test.ts

@@ -231,4 +231,45 @@ describe('executeLocalFilesystemTool', () => {
     })
     expect(mockReportCompletion).not.toHaveBeenCalled()
   })
+
+  it('resolves only once the tool has settled and its result is reported', async () => {
+    let finishRead: (response: unknown) => void = () => {}
+    let finishReport: () => void = () => {}
+    localFilesystem.mockImplementation(async (request: { operation: string }) => {
+      if (request.operation === 'list_mounts') return { ok: true, data: { mounts: [mount] } }
+      return new Promise((resolve) => {
+        finishRead = resolve
+      })
+    })
+    mockReportCompletion.mockImplementation(
+      () =>
+        new Promise<void>((resolve) => {
+          finishReport = resolve
+        })
+    )
+    let resolved = false
+    const running = executeLocalFilesystemTool(
+      'tool-held',
+      'read',
+      { path: `${vfsRoot}/README.md` },
+      { workspaceId: 'ws-1' }
+    ).then(() => {
+      resolved = true
+    })
+
+    await vi.waitFor(() =>
+      expect(localFilesystem).toHaveBeenCalledWith(expect.objectContaining({ operation: 'read' }))
+    )
+    await new Promise((resolve) => setTimeout(resolve, 10))
+    expect(resolved).toBe(false)
+
+    finishRead({ ok: true, data: { content: 'hello', totalLines: 1 } })
+    await vi.waitFor(() => expect(mockReportCompletion).toHaveBeenCalled())
+    await new Promise((resolve) => setTimeout(resolve, 10))
+    expect(resolved).toBe(false)
+
+    finishReport()
+    await running
+    expect(resolved).toBe(true)
+  })
 })