- Home
- Skills
- Testing & Debugging
- Focus-Safe Windows UI Automation Test Harness (.NET)
Focus-Safe Windows UI Automation Test Harness (.NET)
Drive the real packaged EXE through UI Automation patterns only, with an isolated data directory, WS_EX_NOACTIVATE required, a WinEvent observer that fails the test on any foreground steal, state-based waits, evidence on hung close, and a private-desktop runner for tray tests
$29
Works with the AI Agents you already use
Focus-Safe Windows UI Automation Test Harness (.NET)
Example session with this skill installed
Our WPF app has a tray mode and runs on developers' machines all day. Write UI tests that drive the packaged EXE without ever stealing focus from the person using the PC, and prove it.
- Read your context and instructions
- Compiled the focus-safe windows ui
Summary
Harness referenced, four tests written, one app change required before the first run.
What I did
- Built
UiHarness(dotnet build src/UiHarness.csproj -c Release: 0 warnings, 0 errors) and referenced it fromAcme.UiTests. - Wrote a
HarnessConfigforAcme Desktop: view map for Home, Settings, Activity log, Help; flags--data-dir,--local-only,--no-splash,--no-activate. - Tests: GUI subsystem and quiet startup with an every-screen walk; settings toggle round trip; a second background launch that must not change the window; headless mode creating no windows.
- Ran them with
APP_UNDER_TESTpointing at the packaged EXE.
Result
Launch failed on purpose: "The isolated app window must have WS_EX_NOACTIVATE when launched with the no-activate flag." Your --no-activate flag hides the splash but does not set the extended style, so the window could still take focus. Set WS_EX_NOACTIVATE in the main window's CreateParams when the flag is present and skip Activate().
After that fix all four tests pass and focus-observation.json records foregroundEvents: 0 for each. The tray test is marked for the private desktop: run it with pwsh scripts/Run-OnPrivateDesktop.ps1 -ExePath <exe> -Command 'dotnet test Acme.UiTests --filter Category=Tray'; the worker writes private-desktop-evidence.json with switchedDesktop: false.
Connects securely to your tools. The creator never sees your data.
About this skill
UI suites that click coordinates break on layout changes; suites that send input steal the keyboard from whoever is at the machine; suites that pass on a debug build miss what the packaged app does. This harness drives the real executable through UI Automation patterns only, keeps its window from ever activating, proves it, and moves the few tests that need activation onto a private desktop nobody sees.
AppDriver (C#, .NET 8 Windows, no packages): launches APP_UNDER_TEST with a fresh data directory under %TEMP% and your local-only, no-splash and no-activate flags; waits for the main window; requires WS_EX_NOACTIVATE; finds by AutomationId or Name; acts through Invoke, SelectionItem, Toggle, Value and ExpandCollapse patterns (never SendInput, mouse or keyboard); Nav waits for the view-name element so a SetText cannot land on a control the next screen replaces; WaitIdle judges off-screen visibility by bounding rectangle; Close returns the exit code or writes window states to close-failure.json and kills the tree after 60 seconds; VisibleWindows(pid) for headless modes.
ForegroundWatch: out-of-process WinEvent hook on EVENT_SYSTEM_FOREGROUND filtered to the app's process; writes focus-observation.json on dispose and throws if the app ever took focus, with private-desktop activations recorded separately.
PrivateDesktop + Run-OnPrivateDesktop.ps1: CreateDesktop in the interactive window station, CreateProcess with lpDesktop for the worker, no SwitchDesktop or input APIs anywhere; the worker refuses to run on the input desktop and writes private-desktop-evidence.json.
Reference: the design principles, what the app must expose, and the evidence files. Example xunit tests included: GUI subsystem check, quiet startup and every-screen walk, settings round trip, quiet second launch, headless mode, tray minimize/close/restore. Builds with dotnet build; no network.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.
Windows only. Requires interactive session. Cannot see custom controls lacking Automation Peers. Example tests require adaptation to your specific app IDs and flags.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 18 days ago
- Software
- dotnet pwsh windows
Permissions
Starts the executable you name with an isolated data directory under %TEMP%, drives it through UI Automation, and writes evidence JSON files there. The private-desktop runner creates a Windows desktop object and a worker process. No network; reads APP_UNDER_TEST and UI_HARNESS_PRIVATE_DESKTOP only.