deepseek-computer-use 0.1.2 → 0.2.0

package/native/Driver.cs

@@ -1,1906 +1,1906 @@
-
-using System;
-using System.Collections.Generic;
-using System.Diagnostics;
-using System.Drawing;
-using System.Drawing.Imaging;
-using System.IO;
-using System.Runtime.InteropServices;
-using System.Text;
-using System.Windows.Automation;
-
-// ---------------------------------------------------------------------------
-// Platform driver contract. Windows is the only implementation today; the
-// interface exists so macOS (AX + CGEvent) and Linux (AT-SPI) drivers can be
-// added without touching the DSH plugin layer or the stdio protocol.
-// ---------------------------------------------------------------------------
-public interface IPlatformDriver : IDisposable
-{
-    string PlatformName { get; }
-    Dictionary<string, object> Ping(Dictionary<string, object> p);
-    Dictionary<string, object> Hello(Dictionary<string, object> p);
-    Dictionary<string, object> RestoreCursor(Dictionary<string, object> p);
-    void Maintenance();
-    Dictionary<string, object> Close(Dictionary<string, object> p);
-    Dictionary<string, object> ListWindows(Dictionary<string, object> p);
-    Dictionary<string, object> ListApps(Dictionary<string, object> p);
-    Dictionary<string, object> GetWindow(Dictionary<string, object> p);
-    Dictionary<string, object> LaunchApp(Dictionary<string, object> p);
-    Dictionary<string, object> GetWindowState(Dictionary<string, object> p);
-    Dictionary<string, object> Click(Dictionary<string, object> p);
-    Dictionary<string, object> PressKey(Dictionary<string, object> p);
-    Dictionary<string, object> TypeText(Dictionary<string, object> p);
-    Dictionary<string, object> Scroll(Dictionary<string, object> p);
-    Dictionary<string, object> Drag(Dictionary<string, object> p);
-    Dictionary<string, object> SetValue(Dictionary<string, object> p);
-    Dictionary<string, object> PerformSecondaryAction(Dictionary<string, object> p);
-    Dictionary<string, object> ActivateWindow(Dictionary<string, object> p);
-    Dictionary<string, object> OverlayShow(Dictionary<string, object> p);
-    Dictionary<string, object> OverlayHide(Dictionary<string, object> p);
-    Dictionary<string, object> Configure(Dictionary<string, object> p);
-    Dictionary<string, object> ApproveApp(Dictionary<string, object> p);
-    Dictionary<string, object> EndTurn(Dictionary<string, object> p);
-    Dictionary<string, object> InvalidateState(Dictionary<string, object> p);
-    Dictionary<string, object> StartAudioRecording(Dictionary<string, object> p);
-    Dictionary<string, object> StopAudioRecording(Dictionary<string, object> p);
-}
-
-/// Per-request time budget. Codex sends x-oai-cua-request-budget-ms with every call
-/// and the helper honours it; here the budget bounds capture waits and is reported
-/// back when a request overruns so the client and the helper agree on the deadline.
-public static class Budget
-{
-    static long _deadlineTicks;
-    static string _cancelFile;
-
-    public static void Begin(long ms, string cancelFile)
-    {
-        _deadlineTicks = ms > 0 ? DateTime.UtcNow.Ticks + ms * TimeSpan.TicksPerMillisecond : 0;
-        _cancelFile = cancelFile;
-    }
-
-    public static void Check()
-    {
-        if (_cancelFile != null && File.Exists(_cancelFile))
-        {
-            string reason = "cancelled";
-            try { reason = File.ReadAllText(_cancelFile).Trim(); } catch { }
-            if (reason != "turn_ended" && reason != "timeout") reason = "cancelled";
-            throw new ToolError(reason, "the native request was stopped; take a fresh observation before continuing");
-        }
-        if (Expired()) throw new ToolError("timeout", "the native request exceeded its time budget");
-    }
-
-    public static long RemainingMs()
-    {
-        if (_deadlineTicks == 0) return long.MaxValue / 4;
-        return (_deadlineTicks - DateTime.UtcNow.Ticks) / TimeSpan.TicksPerMillisecond;
-    }
-
-    public static bool Expired()
-    {
-        return _deadlineTicks != 0 && RemainingMs() <= 0;
-    }
-}
-
-public class ToolError : Exception
-{
-    public readonly string Code;
-    public ToolError(string code, string message) : base(message) { Code = code; }
-}
-
-// ---------------------------------------------------------------------------
-// Windows driver: Win32 + UI Automation + PrintWindow.
-// ---------------------------------------------------------------------------
-public class WindowsDriver : IPlatformDriver
-{
-    public string PlatformName { get { return "windows"; } }
-
-    // ---- win32 ----
-    delegate bool EnumWindowsProc(IntPtr hWnd, IntPtr lParam);
-
-    [DllImport("user32.dll")] static extern bool EnumWindows(EnumWindowsProc cb, IntPtr lParam);
-    [DllImport("user32.dll")] static extern bool IsWindowVisible(IntPtr hWnd);
-    [DllImport("user32.dll")] static extern bool IsWindow(IntPtr hWnd);
-    [DllImport("user32.dll", CharSet = CharSet.Unicode)] static extern int GetWindowTextW(IntPtr hWnd, StringBuilder sb, int max);
-    [DllImport("user32.dll", CharSet = CharSet.Unicode)] static extern int GetClassNameW(IntPtr hWnd, StringBuilder sb, int max);
-    [DllImport("user32.dll")] static extern uint GetWindowThreadProcessId(IntPtr hWnd, out uint pid);
-    [DllImport("user32.dll")] static extern bool GetWindowRect(IntPtr hWnd, out RECT r);
-    [DllImport("user32.dll")] static extern bool GetClientRect(IntPtr hWnd, out RECT r);
-    [DllImport("user32.dll")] static extern bool PrintWindow(IntPtr hWnd, IntPtr hdc, uint flags);
-    [DllImport("user32.dll")] static extern bool SetForegroundWindow(IntPtr hWnd);
-    [DllImport("user32.dll")] static extern bool ShowWindow(IntPtr hWnd, int cmd);
-    [DllImport("user32.dll")] static extern bool IsIconic(IntPtr hWnd);
-    [DllImport("user32.dll")] static extern bool SetCursorPos(int x, int y);
-    [DllImport("user32.dll")] static extern bool GetCursorPos(out POINT p);
-    [DllImport("user32.dll")] static extern uint SendInput(uint n, INPUT[] inputs, int size);
-    [DllImport("user32.dll")] static extern IntPtr GetForegroundWindow();
-    [DllImport("user32.dll")] static extern IntPtr SetFocus(IntPtr hWnd);
-    [DllImport("user32.dll")] static extern bool AttachThreadInput(uint idAttach, uint idAttachTo, bool fAttach);
-    [DllImport("kernel32.dll")] static extern uint GetCurrentThreadId();
-    [DllImport("user32.dll")] static extern uint GetWindowThreadProcessId(IntPtr hWnd, IntPtr pid);
-    [DllImport("user32.dll")] static extern bool IsWindowEnabled(IntPtr hWnd);
-    [DllImport("user32.dll")] static extern bool SystemParametersInfo(uint action, uint param, IntPtr pv, uint winIni);
-    [DllImport("user32.dll")] static extern bool SwitchToThisWindow(IntPtr hWnd, bool fAltTab);
-    [DllImport("user32.dll")] static extern void keybd_event(byte bVk, byte bScan, uint dwFlags, UIntPtr dwExtraInfo);
-    [DllImport("user32.dll")] static extern bool BringWindowToTop(IntPtr hWnd);
-    [DllImport("user32.dll")] static extern IntPtr SetThreadDpiAwarenessContext(IntPtr ctx);
-    [DllImport("user32.dll")] static extern int GetSystemMetrics(int index);
-    [DllImport("user32.dll")] static extern bool SetProcessDPIAware();
-    [DllImport("dwmapi.dll")] static extern int DwmGetWindowAttribute(IntPtr hwnd, int attr, out RECT r, int size);
-    [DllImport("user32.dll")] static extern IntPtr GetWindow(IntPtr hWnd, uint cmd);
-    [DllImport("user32.dll")] static extern int GetWindowLong(IntPtr hWnd, int index);
-
-    [StructLayout(LayoutKind.Sequential)] struct RECT { public int Left, Top, Right, Bottom; }
-    [StructLayout(LayoutKind.Sequential)] struct POINT { public int X, Y; }
-
-    [StructLayout(LayoutKind.Sequential)]
-    struct MOUSEINPUT { public int dx; public int dy; public uint mouseData; public uint dwFlags; public uint time; public IntPtr dwExtraInfo; }
-
-    [StructLayout(LayoutKind.Sequential)]
-    struct KEYBDINPUT { public ushort wVk; public ushort wScan; public uint dwFlags; public uint time; public IntPtr dwExtraInfo; }
-
-    [StructLayout(LayoutKind.Explicit)]
-    struct INPUTUNION
-    {
-        [FieldOffset(0)] public MOUSEINPUT mi;
-        [FieldOffset(0)] public KEYBDINPUT ki;
-    }
-
-    [StructLayout(LayoutKind.Sequential)]
-    struct INPUT { public uint type; public INPUTUNION u; }
-
-    const int INPUT_MOUSE = 0;
-    const int INPUT_KEYBOARD = 1;
-    const uint KEYEVENTF_KEYUP = 0x0002;
-    const uint KEYEVENTF_UNICODE = 0x0004;
-    const uint KEYEVENTF_EXTENDEDKEY = 0x0001;
-    const uint MOUSEEVENTF_LEFTDOWN = 0x0002;
-    const uint MOUSEEVENTF_LEFTUP = 0x0004;
-    const uint MOUSEEVENTF_RIGHTDOWN = 0x0008;
-    const uint MOUSEEVENTF_RIGHTUP = 0x0010;
-    const uint MOUSEEVENTF_MIDDLEDOWN = 0x0020;
-    const uint MOUSEEVENTF_MIDDLEUP = 0x0040;
-    const uint MOUSEEVENTF_WHEEL = 0x0800;
-    const uint MOUSEEVENTF_HWHEEL = 0x1000;
-    const int DWMWA_EXTENDED_FRAME_BOUNDS = 9;
-
-    // ---- state ----
-    readonly object _sync = new object();
-    sealed class Snapshot
-    {
-        public long WindowId;
-        public DateTime At;
-        public Dictionary<long, AutomationElement> Elements = new Dictionary<long, AutomationElement>();
-        public Dictionary<long, string> FallbackLabels = new Dictionary<long, string>();
-        public List<object> Summary = new List<object>();
-    }
-
-    readonly Dictionary<long, Snapshot> _snapshots = new Dictionary<long, Snapshot>();
-    int _snapshotTtlMs = 120000;
-
-    /// A screenshot region the model was handed, so a later coordinate action can
-    /// be interpreted in that exact screenshot space (Codex: screenshotId).
-    sealed class ShotRegion
-    {
-        public long WindowId;
-        public int OriginX;
-        public int OriginY;
-        public int Width;
-        public int Height;
-        /// Physical pixels per screenshot pixel. 1.0 for DPI-aware targets; >1 when
-        /// the target is DPI-virtualized and PrintWindow returns an unscaled render.
-        public double ScaleX = 1.0;
-        public double ScaleY = 1.0;
-        public DateTime At;
-    }
-
-    // In-use indicator ("DSH 正在使用您的电脑" + whale-girl). Codex draws an animated
-    // border and a virtual cursor; this is the DSH equivalent, and Esc is the kill
-    // switch: the overlay drops a cancel file that every input path checks first.
-    Process _overlay;
-    System.Threading.Timer _cursorWatchdog;
-    DateTime _overlayUntil = DateTime.MinValue;
-    string _overlayCancelFile;
-    string _overlayText = "DSH 正在使用您的电脑";
-    string _overlayHint = "Esc 取消";
-    string _overlayAccent = "#3f7fd8";
-    string _cursorFile;      // the overlay polls this to draw the virtual cursor
-    /// Turn token that the user interrupted with Esc. Codex persists the same
-    /// fact as a marker file per conversation+turn; here the token travels with
-    /// every input call, so the block ends exactly when the turn does.
-    string _interruptedTurn = null;
-    /// Apps the plugin approved through the DSH approval seam. Codex passes the same
-    /// fact per request as x-oai-cua-approved-app; here the helper is the enforcer.
-    readonly HashSet<string> _approvedApps = new HashSet<string>(StringComparer.OrdinalIgnoreCase);
-    bool _requireAppApproval = false;
-    Process _audioProcess;
-    string _audioPath;
-    string _audioStopFile;
-
-    readonly Dictionary<string, ShotRegion> _shots = new Dictionary<string, ShotRegion>();
-    int _shotTtlMs = 120000;
-    int _shotSeq = 0;
-    readonly string _shotGeneration = Guid.NewGuid().ToString("N");
-
-    const uint GW_OWNER = 4;
-    const int GWL_STYLE = -16;
-    const int WS_POPUP = unchecked((int)0x80000000);
-
-    public WindowsDriver()
-    {
-        Maintenance(); // startup repair: a previous session may have died with cursors hidden
-        try { SetThreadDpiAwarenessContext(new IntPtr(-4)); } catch { }
-        try { SetProcessDPIAware(); } catch { }
-    }
-
-    public void Dispose()
-    {
-        if (_cursorWatchdog != null) { _cursorWatchdog.Dispose(); _cursorWatchdog = null; }
-        StopOverlay();
-    }
-
-    // ---- approval plumbing ----
-    public Dictionary<string, object> Configure(Dictionary<string, object> p)
-    {
-        _requireAppApproval = Json.GetBool(p, "requireAppApproval", _requireAppApproval);
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["requireAppApproval"] = _requireAppApproval;
-        return r;
-    }
-
-    public Dictionary<string, object> ApproveApp(Dictionary<string, object> p)
-    {
-        string app = Json.GetString(p, "app");
-        if (string.IsNullOrEmpty(app)) throw new ToolError("bad_request", "approve_app requires 'app'");
-        lock (_sync) { _approvedApps.Add(app); }
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["approved"] = app;
-        r["count"] = (long)_approvedApps.Count;
-        return r;
-    }
-
-    /// Turn boundary: drop per-turn state (element snapshots, screenshot regions, the
-    /// Esc block, the indicator) but keep session-scoped app approvals.
-    public Dictionary<string, object> EndTurn(Dictionary<string, object> p)
-    {
-        string ending = Json.GetString(p, "turnToken");
-        if (ending != null) ClearInterrupt(ending);
-        lock (_sync) { _snapshots.Clear(); _shots.Clear(); }
-        _interruptedTurn = null;
-        StopOverlay();
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["ended"] = true;
-        r["approvedApps"] = (long)_approvedApps.Count;
-        return r;
-    }
-
-    public Dictionary<string, object> InvalidateState(Dictionary<string, object> p)
-    {
-        lock (_sync) { _snapshots.Clear(); _shots.Clear(); }
-        StopOverlay();
-        return new Dictionary<string, object> { { "invalidated", true } };
-    }
-
-    void AssertAppApproved(IntPtr h)
-    {
-        if (!_requireAppApproval) return;
-        uint pid;
-        GetWindowThreadProcessId(h, out pid);
-        string id = ProcessPathOf(pid);
-        lock (_sync) { if (_approvedApps.Contains(id)) return; }
-        throw new ToolError("app_not_approved", "the Computer Use policy has not approved '" + id + "' yet. Approve the app through the DSH approval prompt before acting on its windows.");
-    }
-
-    // ---- audio recording (WASAPI loopback) ----
-    public Dictionary<string, object> StartAudioRecording(Dictionary<string, object> p)
-    {
-        bool envEnabled = Environment.GetEnvironmentVariable("DSH_CU_AUDIO") == "1";
-        bool force = Json.GetBool(p, "force", false);
-        // Codex treats computer audio as its own app ("computer-audio") and asks for a
-        // separate approval before recording; the plugin passes that here.
-        bool audioApproved = Json.GetString(p, "approvedApp") == "computer-audio";
-        if (!envEnabled && !force && !audioApproved)
-            throw new ToolError("audio_not_enabled", "computer audio recording is not enabled. Approve it (app \"computer-audio\") or set DSH_CU_AUDIO=1.");
-        if (_audioProcess != null && !_audioProcess.HasExited)
-            throw new ToolError("audio_running", "an audio recording is already running; stop it first");
-
-        string dir = Path.GetDirectoryName(typeof(WindowsDriver).Assembly.Location) ?? ".";
-        string exe = Path.Combine(dir, "audio", "dsh-cu-audio.exe");
-        if (!File.Exists(exe)) throw new ToolError("audio_missing", "audio recorder not found at " + exe);
-
-        long maxMs = Json.GetLong(p, "maxDurationMs", 60000);
-        if (maxMs < 500) maxMs = 500;
-        if (maxMs > 300000) maxMs = 300000;   // Codex caps a single recording at 5 minutes
-        _audioPath = Path.Combine(Path.GetTempPath(), "dsh-cu-audio-" + Guid.NewGuid().ToString("N") + ".wav");
-        _audioStopFile = Path.Combine(Path.GetTempPath(), "dsh-cu-audio-stop-" + Guid.NewGuid().ToString("N") + ".flag");
-        try { if (File.Exists(_audioStopFile)) File.Delete(_audioStopFile); } catch { }
-
-        ProcessStartInfo psi = new ProcessStartInfo(exe, "--out \"" + _audioPath + "\" --max-ms " + maxMs + " --stop-file \"" + _audioStopFile + "\"");
-        psi.UseShellExecute = false;
-        psi.CreateNoWindow = true;
-        psi.RedirectStandardOutput = true;
-        psi.RedirectStandardError = true;
-        _audioProcess = Process.Start(psi);
-        System.Threading.Thread.Sleep(400);
-        if (_audioProcess.HasExited)
-            throw new ToolError("audio_not_started", "computer audio recording did not start in time (the recorder exited immediately)");
-
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["started"] = true;
-        r["filepath"] = _audioPath;
-        r["maxDurationMs"] = maxMs;
-        return r;
-    }
-
-    public Dictionary<string, object> StopAudioRecording(Dictionary<string, object> p)
-    {
-        if (_audioProcess == null) throw new ToolError("audio_not_started", "no audio recording is running");
-        try { if (_audioStopFile != null) File.WriteAllText(_audioStopFile, "stop"); } catch { }
-        try { _audioProcess.WaitForExit(8000); } catch { }
-        try { if (_audioProcess != null && !_audioProcess.HasExited) _audioProcess.Kill(); } catch { }
-
-        string path = _audioPath;
-        _audioProcess = null;
-        _audioStopFile = null;
-        _audioPath = null;
-
-        byte[] audio = new byte[0];
-        try { if (path != null && File.Exists(path)) audio = File.ReadAllBytes(path); } catch { }
-
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["filepath"] = path;
-        // Codex returns the raw WAV bytes plus a data URL; base64 is the JSON-safe form.
-        r["bytes"] = Convert.ToBase64String(audio);
-        r["byteLength"] = (long)audio.Length;
-        r["data_url"] = audio.Length > 44 ? "data:audio/wav;base64," + Convert.ToBase64String(audio) : "";
-        r["ok"] = audio.Length > 44;
-        if (audio.Length <= 44) throw new ToolError("audio_not_recorded", "computer audio recording produced no samples");
-        return r;
-    }
-
-    // ---- in-use indicator ----
-    static string OverlayExePath()
-    {
-        string dir = Path.GetDirectoryName(typeof(WindowsDriver).Assembly.Location);
-        return Path.Combine(dir ?? ".", "dsh-cu-overlay.exe");
-    }
-
-    public Dictionary<string, object> OverlayShow(Dictionary<string, object> p)
-    {
-        string text = Json.GetString(p, "text");
-        if (!string.IsNullOrEmpty(text)) _overlayText = text;
-        string hint = Json.GetString(p, "hint");
-        if (!string.IsNullOrEmpty(hint)) _overlayHint = hint;
-        string accent = Json.GetString(p, "accent");
-        if (!string.IsNullOrEmpty(accent)) _overlayAccent = accent;
-        EnsureOverlay(Json.GetLong(p, "ttlMs", 20000));
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["visible"] = true;
-        r["cancelFile"] = _overlayCancelFile;
-        return r;
-    }
-
-    public Dictionary<string, object> OverlayHide(Dictionary<string, object> p)
-    {
-        StopOverlay();
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["visible"] = false;
-        return r;
-    }
-
-    void EnsureOverlay(long ttlMs)
-    {
-        string exe = OverlayExePath();
-        if (!File.Exists(exe)) return;
-        string mode = Environment.GetEnvironmentVariable("DSH_CU_INDICATOR");
-        if (mode != null && mode.Equals("off", StringComparison.OrdinalIgnoreCase)) { StopOverlay(); return; }
-        string ttlEnv = Environment.GetEnvironmentVariable("DSH_CU_INDICATOR_TTL_MS");
-        long parsed;
-        if (ttlEnv != null && long.TryParse(ttlEnv, out parsed) && parsed >= 1000) ttlMs = parsed;
-        if (_overlay != null && !_overlay.HasExited && DateTime.UtcNow < _overlayUntil)
-        {
-            _overlayUntil = DateTime.UtcNow.AddMilliseconds(ttlMs);
-            return;
-        }
-        StopOverlay();
-        _overlayCancelFile = Path.Combine(Path.GetTempPath(), "dsh-cu-cancel-" + Guid.NewGuid().ToString("N") + ".flag");
-        if (_cursorFile == null) _cursorFile = Path.Combine(Path.GetTempPath(), "dsh-cu-cursor-" + Guid.NewGuid().ToString("N").Substring(0, 8) + ".flag");
-        try { if (File.Exists(_overlayCancelFile)) File.Delete(_overlayCancelFile); } catch { }
-        StringBuilder args = new StringBuilder();
-        args.Append("--ttl-ms ").Append(ttlMs);
-        args.Append(" --cancel-file \"").Append(_overlayCancelFile).Append('"');
-        args.Append(" --text \"").Append(_overlayText).Append('"');
-        args.Append(" --hint \"").Append(_overlayHint).Append('"');
-        args.Append(" --owner-pid ").Append(Process.GetCurrentProcess().Id);
-        args.Append(" --accent \"").Append(_overlayAccent).Append('"');
-        if (_cursorFile != null) args.Append(" --cursor-file \"").Append(_cursorFile).Append('"');
-        // Suppression is the default (the virtual cursor stands in for the real one);
-        // DSH_CU_KEEP_CURSOR=1 keeps the real pointer visible instead.
-        if (Environment.GetEnvironmentVariable("DSH_CU_KEEP_CURSOR") == "1") args.Append(" --no-suppress-cursor");
-        try
-        {
-            ProcessStartInfo psi = new ProcessStartInfo(exe, args.ToString());
-            psi.UseShellExecute = false;
-            _overlay = Process.Start(psi);
-            _overlayUntil = DateTime.UtcNow.AddMilliseconds(ttlMs);
-            // Codex runs a dedicated system-cursor-manager process that restores the
-            // pointer the moment its parent dies. This timer gives the same guarantee
-            // without a second exe: if the overlay vanishes with cursors hidden, the
-            // pointer comes back within half a second, with no request needed.
-            if (_cursorWatchdog == null)
-                _cursorWatchdog = new System.Threading.Timer(delegate { try { Maintenance(); } catch { } }, null, 500, 500);
-        }
-        catch { _overlay = null; }
-    }
-
-    void StopOverlay()
-    {
-        try { if (_overlay != null && !_overlay.HasExited) _overlay.Kill(); } catch { }
-        // Kill() skips the overlay's exit handlers, which is where it would restore the
-        // system cursors. Restore them from here so a terminated overlay can never leave
-        // the user without a pointer.
-        RestoreCursors();
-        try { File.Delete(OverlayCursorFlagPath()); } catch { }
-        try { if (_overlayCancelFile != null && File.Exists(_overlayCancelFile)) File.Delete(_overlayCancelFile); } catch { }
-        _overlay = null;
-        _overlayUntil = DateTime.MinValue;
-    }
-
-    /// Esc on the overlay drops a cancel file; honour it before any other input.
-    const string EscMessage = "Computer Use was stopped by the user with the physical Escape key. Stop your work, do not call further Computer Use tools in this turn, and send a final message noting that the user stopped Computer Use.";
-
-    void CheckCancel(Dictionary<string, object> p)
-    {
-        Budget.Check();
-        string token = p == null ? null : Json.GetString(p, "turnToken");
-        if (token != null && (token == _interruptedTurn || InterruptMarkerExists(token)))
-            throw new ToolError("turn_stopped", EscMessage);
-        bool overlayExited = _overlay != null && _overlay.HasExited && _overlay.ExitCode == 130;
-        bool cancelFile = _overlayCancelFile != null && File.Exists(_overlayCancelFile);
-        if (!overlayExited && !cancelFile) return;
-        if (cancelFile) { try { File.Delete(_overlayCancelFile); } catch { } }
-        _interruptedTurn = token;
-        PersistInterrupt(token);
-        StopOverlay();
-        Program.RequestedExitCode = 130;   // Codex exits 130 when the user hits Esc
-        throw new ToolError("cancelled_by_user", EscMessage);
-    }
-
-    const int ProtocolVersion = 2;
-
-    static string InterruptMarkerPath(string token)
-    {
-        string root = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "dsh-computer-use", "interrupts");
-        StringBuilder safe = new StringBuilder();
-        if (token != null)
-        {
-            for (int i = 0; i < token.Length && safe.Length < 64; i++)
-            {
-                char ch = token[i];
-                if (char.IsLetterOrDigit(ch) || ch == '-' || ch == '_' || ch == '.') safe.Append(ch);
-                else if (ch == ':') safe.Append('-');
-            }
-        }
-        if (safe.Length == 0) safe.Append("turn");
-        return Path.Combine(root, safe.ToString() + ".flag");
-    }
-
-    static void PersistInterrupt(string token)
-    {
-        try
-        {
-            string path = InterruptMarkerPath(token);
-            Directory.CreateDirectory(Path.GetDirectoryName(path));
-            File.WriteAllText(path, DateTime.UtcNow.ToString("o"));
-        }
-        catch { }
-    }
-
-    static bool InterruptMarkerExists(string token)
-    {
-        try { return File.Exists(InterruptMarkerPath(token)); } catch { return false; }
-    }
-
-    static void ClearInterrupt(string token)
-    {
-        try { File.Delete(InterruptMarkerPath(token)); } catch { }
-    }
-
-    /// Publish the pointer target for the virtual cursor. The overlay reads this
-    /// file; a missing file simply means "no cursor to draw".
-    void MarkCursor(int x, int y)
-    {
-        if (_cursorFile == null) return;
-        try { File.WriteAllText(_cursorFile, x.ToString(System.Globalization.CultureInfo.InvariantCulture) + "," + y.ToString(System.Globalization.CultureInfo.InvariantCulture) + "," + DateTime.UtcNow.Ticks.ToString(System.Globalization.CultureInfo.InvariantCulture)); } catch { }
-    }
-
-    /// Codex reports screenLocked instead of failing with an opaque automation error.
-    static void AssertUnlocked()
-    {
-        IntPtr fg = GetForegroundWindow();
-        if (fg == IntPtr.Zero) return;
-        uint pid;
-        GetWindowThreadProcessId(fg, out pid);
-        string proc = "";
-        try { proc = Process.GetProcessById((int)pid).ProcessName; } catch { }
-        if (proc.Equals("LockApp", StringComparison.OrdinalIgnoreCase) || proc.Equals("LogonUI", StringComparison.OrdinalIgnoreCase))
-            throw new ToolError("screen_locked", "the Windows desktop is locked (LockApp is in the foreground), so Computer Use cannot interact with it. Ask the user to unlock the desktop and stop trying until they do.");
-    }
-
-    // ---- helpers ----
-    static string TitleOf(IntPtr h)
-    {
-        StringBuilder sb = new StringBuilder(1024);
-        GetWindowTextW(h, sb, sb.Capacity);
-        return sb.ToString();
-    }
-
-    static string ClassOf(IntPtr h)
-    {
-        StringBuilder sb = new StringBuilder(256);
-        GetClassNameW(h, sb, sb.Capacity);
-        return sb.ToString();
-    }
-
-    static RECT BoundsOf(IntPtr h)
-    {
-        RECT r;
-        try
-        {
-            if (DwmGetWindowAttribute(h, DWMWA_EXTENDED_FRAME_BOUNDS, out r, Marshal.SizeOf(typeof(RECT))) == 0 && r.Right > r.Left)
-                return r;
-        }
-        catch { }
-        GetWindowRect(h, out r);
-        return r;
-    }
-
-    static Dictionary<string, object> RectDict(int x, int y, int w, int h)
-    {
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["x"] = x; d["y"] = y; d["width"] = w; d["height"] = h;
-        return d;
-    }
-
-    static string ProcessPathOf(uint pid)
-    {
-        try
-        {
-            Process p = Process.GetProcessById((int)pid);
-            try { return p.MainModule.FileName; }
-            catch { return p.ProcessName; }
-        }
-        catch { return "unknown"; }
-    }
-
-    static string ProcessNameOf(uint pid)
-    {
-        try { return Process.GetProcessById((int)pid).ProcessName; }
-        catch { return "unknown"; }
-    }
-
-    Dictionary<string, object> AppOf(uint pid)
-    {
-        Dictionary<string, object> app = new Dictionary<string, object>();
-        string full = ProcessPathOf(pid);
-        app["id"] = full;
-        app["name"] = Path.GetFileNameWithoutExtension(full);
-        app["displayName"] = ProcessNameOf(pid);
-        app["pid"] = (long)pid;
-        try
-        {
-            Process proc = Process.GetProcessById((int)pid);
-            app["lastUsedDate"] = proc.StartTime.ToUniversalTime().ToString("o");
-        }
-        catch { }
-        return app;
-    }
-
-    Dictionary<string, object> WindowDict(IntPtr h, uint pid, int z)
-    {
-        RECT r = BoundsOf(h);
-        Dictionary<string, object> w = new Dictionary<string, object>();
-        w["id"] = (long)h.ToInt64();
-        w["title"] = TitleOf(h);
-        w["app"] = AppOf(pid);
-        w["bounds"] = RectDict(r.Left, r.Top, r.Right - r.Left, r.Bottom - r.Top);
-        w["zIndex"] = z;
-        w["className"] = ClassOf(h);
-        return w;
-    }
-
-    List<IntPtr> EnumerateWindows(bool visibleOnly)
-    {
-        List<IntPtr> list = new List<IntPtr>();
-        EnumWindows(delegate(IntPtr h, IntPtr lp)
-        {
-            if (!IsWindow(h)) return true;
-            if (visibleOnly && !IsWindowVisible(h)) return true;
-            uint pid;
-            GetWindowThreadProcessId(h, out pid);
-            if (pid == 0) return true;
-            RECT r;
-            GetWindowRect(h, out r);
-            if (r.Right - r.Left < 80 || r.Bottom - r.Top < 60) return true;
-            string title = TitleOf(h);
-            if (title.Length == 0) return true;
-            list.Add(h);
-            return true;
-        }, IntPtr.Zero);
-        return list;
-    }
-
-    // ---- methods ----
-    public Dictionary<string, object> Ping(Dictionary<string, object> p)
-    {
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["platform"] = PlatformName;
-        d["pid"] = (long)Process.GetCurrentProcess().Id;
-        d["overlayPid"] = _overlay != null && !_overlay.HasExited ? (object)(long)_overlay.Id : null;
-        d["ok"] = true;
-        d["uia"] = true;
-        return d;
-    }
-
-    /// Run cheaply before every request: if a previous overlay died while it was hiding
-    /// the system cursors, put them back. A blank desktop cursor is worse than a stray
-    /// visible one, so this never assumes the overlay cleaned up after itself.
-    public void Maintenance()
-    {
-        try
-        {
-            string flag = OverlayCursorFlagPath();
-            if (!File.Exists(flag)) return;
-            int pid;
-            if (!int.TryParse(File.ReadAllText(flag).Trim(), out pid)) { RestoreCursors(); return; }
-            bool alive;
-            try { Process.GetProcessById(pid); alive = true; } catch { alive = false; }
-            if (!alive) { RestoreCursors(); try { File.Delete(flag); } catch { } }
-        }
-        catch { }
-    }
-
-    static string OverlayCursorFlagPath()
-    {
-        return Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "dsh-computer-use", "cursor-hidden.flag");
-    }
-
-    static void RestoreCursors()
-    {
-        try { SystemParametersInfo(0x0057, 0, IntPtr.Zero, 0); } catch { }
-    }
-
-    public Dictionary<string, object> RestoreCursor(Dictionary<string, object> p)
-    {
-        RestoreCursors();
-        try { File.Delete(OverlayCursorFlagPath()); } catch { }
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["restored"] = true;
-        return r;
-    }
-
-    public Dictionary<string, object> Hello(Dictionary<string, object> p)
-    {
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["ok"] = true;
-        d["name"] = "dsh-computer-use";
-        d["protocolVersion"] = ProtocolVersion;
-        d["platform"] = PlatformName;
-        List<object> caps = new List<object>();
-        foreach (string c in new string[] {
-            "window2", "accessibility_tree", "multi_screenshot", "wgc_capture",
-            "virtual_cursor", "audio_recording", "turn_interrupt", "app_approval",
-            "screenshot_ids", "budgeted_capture" }) caps.Add(c);
-        d["capabilities"] = caps;
-        return d;
-    }
-
-    public Dictionary<string, object> Close(Dictionary<string, object> p)
-    {
-        Program.RequestedExitCode = 0;
-        Program.CloseRequested = true;
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["closed"] = true;
-        return d;
-    }
-
-    public Dictionary<string, object> ListWindows(Dictionary<string, object> p)
-    {
-        List<object> outList = new List<object>();
-        List<IntPtr> hs = EnumerateWindows(true);
-        for (int i = 0; i < hs.Count; i++)
-        {
-            uint pid;
-            GetWindowThreadProcessId(hs[i], out pid);
-            outList.Add(WindowDict(hs[i], pid, i));
-        }
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["windows"] = outList;
-        d["count"] = outList.Count;
-        return d;
-    }
-
-    public Dictionary<string, object> ListApps(Dictionary<string, object> p)
-    {
-        Dictionary<string, List<object>> grouped = new Dictionary<string, List<object>>();
-        Dictionary<string, Dictionary<string, object>> appInfo = new Dictionary<string, Dictionary<string, object>>();
-        List<IntPtr> hs = EnumerateWindows(true);
-        for (int i = 0; i < hs.Count; i++)
-        {
-            uint pid;
-            GetWindowThreadProcessId(hs[i], out pid);
-            Dictionary<string, object> app = AppOf(pid);
-            string key = Convert.ToString(app["id"]);
-            if (!grouped.ContainsKey(key)) { grouped[key] = new List<object>(); appInfo[key] = app; }
-            grouped[key].Add(WindowDict(hs[i], pid, i));
-        }
-        List<object> apps = new List<object>();
-        foreach (KeyValuePair<string, List<object>> kv in grouped)
-        {
-            Dictionary<string, object> a = appInfo[kv.Key];
-            a["isRunning"] = true;
-            try
-            {
-                IntPtr sample = IntPtr.Zero;
-                List<IntPtr> windowsOfApp = new List<IntPtr>();
-                for (int w = 0; w < hs.Count; w++)
-                {
-                    uint wpid;
-                    GetWindowThreadProcessId(hs[w], out wpid);
-                    if (Convert.ToString(AppOf(wpid)["id"]) == kv.Key) { sample = hs[w]; break; }
-                }
-                if (sample != IntPtr.Zero)
-                {
-                    uint spid;
-                    GetWindowThreadProcessId(sample, out spid);
-                    a["lastUsedDate"] = Process.GetProcessById((int)spid).StartTime.ToUniversalTime().ToString("o");
-                }
-            }
-            catch { }
-            a["windows"] = kv.Value;
-            apps.Add(a);
-        }
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["apps"] = apps;
-        d["count"] = apps.Count;
-        return d;
-    }
-
-    public Dictionary<string, object> GetWindow(Dictionary<string, object> p)
-    {
-        IntPtr h = ResolveWindow(p);
-        uint pid;
-        GetWindowThreadProcessId(h, out pid);
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["window"] = WindowDict(h, pid, 0);
-        return d;
-    }
-
-    public Dictionary<string, object> ActivateWindow(Dictionary<string, object> p)
-    {
-        IntPtr h = ResolveWindow(p);
-        Activate(h);
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["activated"] = true;
-        d["window"] = WindowDict(h, PidOf(h), 0);
-        return d;
-    }
-
-    static uint PidOf(IntPtr h) { uint pid; GetWindowThreadProcessId(h, out pid); return pid; }
-
-    /// Bring a window to the foreground and PROVE it got there. Input is never
-    /// injected into a window that is not foreground: SendInput targets whatever
-    /// holds focus, so an unproven activation would type into the user's window.
-    static void Activate(IntPtr h)
-    {
-        if (!EnsureForeground(h))
-            throw new ToolError("focus_failed", "could not bring window " + h.ToInt64() + " to the foreground, so the input was refused instead of landing in another window. " + DescribeForeground());
-    }
-
-    static bool EnsureForeground(IntPtr h)
-    {
-        if (!IsWindow(h)) throw new ToolError("window_gone", "window " + h.ToInt64() + " no longer exists; call list_windows again");
-        if (!IsWindowEnabled(h)) throw new ToolError("window_disabled", "window " + h.ToInt64() + " is disabled (a modal dialog may be blocking it)");
-        if (IsIconic(h)) { ShowWindow(h, 9); System.Threading.Thread.Sleep(180); }
-        if (GetForegroundWindow() == h) return true;
-
-        uint fgThread = GetWindowThreadProcessId(GetForegroundWindow(), IntPtr.Zero);
-        uint thisThread = GetCurrentThreadId();
-        bool attached = false;
-        try
-        {
-            if (fgThread != 0 && fgThread != thisThread) attached = AttachThreadInput(thisThread, fgThread, true);
-            ShowWindow(h, 5); // SW_SHOW
-            BringWindowToTop(h);
-            SetForegroundWindow(h);
-            SetFocus(h);
-        }
-        finally { if (attached) AttachThreadInput(thisThread, fgThread, false); }
-
-        if (WaitForeground(h, 10)) return true;
-
-        // Windows refuses foreground changes from a background process. The
-        // documented-looking escape hatches, in order of intrusiveness:
-        //   1. SwitchToThisWindow (alt-tab semantics, no key injection)
-        //   2. a bare Alt press, which lifts the foreground lock for this call
-        try { SwitchToThisWindow(h, true); } catch { }
-        if (WaitForeground(h, 10)) return true;
-
-        keybd_event(0x12, 0, 0, UIntPtr.Zero);        // VK_MENU down
-        keybd_event(0x12, 0, 0x0002, UIntPtr.Zero);   // VK_MENU up
-        SetForegroundWindow(h);
-        if (WaitForeground(h, 12)) return true;
-        return false;
-    }
-
-    /// Name the process that currently owns focus: the usual reason a
-    /// background host cannot steal it is an always-on-top utility or an
-    /// elevated window.
-    static string DescribeForeground()
-    {
-        IntPtr fg = GetForegroundWindow();
-        if (fg == IntPtr.Zero) return "No window currently holds the foreground.";
-        uint pid;
-        GetWindowThreadProcessId(fg, out pid);
-        string proc = "unknown";
-        try { proc = Process.GetProcessById((int)pid).ProcessName; } catch { }
-        string title = TitleOf(fg);
-        string cls = ClassOf(fg);
-        return "Foreground is held by " + proc + " (pid " + pid + ", class " + cls + (title.Length > 0 ? ", title \"" + Trim(title, 60) + "\"" : "") + "). Bring the target window forward yourself, or close/pause the utility that keeps stealing focus.";
-    }
-
-    static bool WaitForeground(IntPtr h, int attempts)
-    {
-        for (int i = 0; i < attempts; i++)
-        {
-            if (GetForegroundWindow() == h) { System.Threading.Thread.Sleep(70); return true; }
-            System.Threading.Thread.Sleep(40);
-        }
-        return GetForegroundWindow() == h;
-    }
-
-    static void Slow() { System.Threading.Thread.Sleep(120); }
-
-    public Dictionary<string, object> LaunchApp(Dictionary<string, object> p)
-    {
-        string app = Json.GetString(p, "app");
-        if (string.IsNullOrEmpty(app)) throw new ToolError("bad_request", "launch_app requires 'app'");
-        // Resolution order matches the API contract: an explicit .exe path first,
-        // then a packaged-app id (AUMID) through shell:AppsFolder, then a shell URI
-        // such as ms-settings:. A bare unknown string is tried verbatim last.
-        string resolved = app;
-        if (!File.Exists(app) && app.IndexOf('\\') < 0 && !app.EndsWith(".exe", StringComparison.OrdinalIgnoreCase) && app.IndexOf(':') < 0)
-            resolved = "shell:AppsFolder\\" + app;
-        ProcessStartInfo psi = new ProcessStartInfo();
-        psi.FileName = resolved;
-        psi.UseShellExecute = true;
-        List<object> args = Json.Get(p, "args") as List<object>;
-        if (args != null && args.Count > 0)
-        {
-            StringBuilder argLine = new StringBuilder();
-            for (int i = 0; i < args.Count; i++)
-            {
-                if (i > 0) argLine.Append(' ');
-                string a = Convert.ToString(args[i], System.Globalization.CultureInfo.InvariantCulture) ?? "";
-                // A quote or newline would let a caller smuggle extra arguments past the
-                // single quoting rule below, so refuse instead of guessing.
-                if (a.IndexOf('"') >= 0 || a.IndexOf('\n') >= 0 || a.IndexOf('\r') >= 0)
-                    throw new ToolError("bad_request", "launch_app argument " + i + " contains a quote or newline; pass a single plain argument");
-                if (a.Length == 0 || a.IndexOf(' ') >= 0 || a.IndexOf('\t') >= 0)
-                    argLine.Append('"').Append(a).Append('"');
-                else argLine.Append(a);
-            }
-            psi.Arguments = argLine.ToString();
-        }
-        Process proc = Process.Start(psi);
-        System.Threading.Thread.Sleep(900);
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["launched"] = true;
-        d["app"] = app;
-        try { d["pid"] = (long)proc.Id; } catch { }
-        return d;
-    }
-
-    IntPtr ResolveWindow(Dictionary<string, object> p)
-    {
-        Dictionary<string, object> w = Json.AsObject(Json.Get(p, "window"));
-        long id = Json.GetLong(w, "id", 0);
-        if (id == 0) id = Json.GetLong(p, "windowId", 0);
-        if (id == 0) throw new ToolError("bad_request", "a 'window' object with a numeric 'id' is required (use list_windows / list_apps)");
-        IntPtr h = new IntPtr(id);
-        if (!IsWindow(h)) throw new ToolError("window_gone", "window " + id + " no longer exists; call list_windows again");
-        return h;
-    }
-
-    // ---- observation ----
-    public Dictionary<string, object> GetWindowState(Dictionary<string, object> p)
-    {
-        IntPtr h = ResolveWindow(p);
-        AssertAppApproved(h);
-        bool includeText = Json.GetBool(p, "include_text", false);
-        bool includeShot = Json.GetBool(p, "include_screenshot", false);
-        long maxElements = Json.GetLong(p, "maxElements", 260);
-        long ttl = Json.GetLong(p, "ttlMs", _snapshotTtlMs);
-
-        uint pid;
-        GetWindowThreadProcessId(h, out pid);
-        Dictionary<string, object> result = new Dictionary<string, object>();
-        result["window"] = WindowDict(h, pid, 0);
-
-        // Codex parity: accessibility is null unless include_text was asked for.
-        result["accessibility"] = includeText ? (object)BuildAccessibility(h, maxElements, (int)ttl) : null;
-        bool blank;
-        result["screenshots"] = includeShot ? (object)CaptureScreenshots(h, out blank) : new List<object>();
-        return result;
-    }
-
-    Dictionary<string, object> BuildAccessibility(IntPtr h, long maxElements, int ttlMs)
-    {
-        Snapshot snap = new Snapshot();
-        snap.WindowId = h.ToInt64();
-        snap.At = DateTime.UtcNow;
-
-        StringBuilder tree = new StringBuilder();
-        int index = 0;
-        AutomationElement root = null;
-        try { root = AutomationElement.FromHandle(h); }
-        catch (Exception ex) { throw new ToolError("uia_unavailable", "UI Automation cannot read this window: " + ex.Message); }
-
-        string focused = null;
-        Walk(root, 0, maxElements, snap, tree, ref index, h, ref focused);
-
-        if (index == 0)
-        {
-            // Electron/Canvas apps may not expose a tree until accessibility is requested;
-            // retry once after a short delay.
-            System.Threading.Thread.Sleep(180);
-            try { root = AutomationElement.FromHandle(h); } catch { }
-            Walk(root, 0, maxElements, snap, tree, ref index, h, ref focused);
-        }
-
-        lock (_sync) { _snapshots[snap.WindowId] = snap; }
-
-        Dictionary<string, object> state = new Dictionary<string, object>();
-        state["tree"] = tree.ToString();
-        state["elementCount"] = (long)index;
-        state["elements"] = snap.Summary;
-
-        // Structured context Codex exposes beside the tree. Each is best-effort:
-        // an app that does not implement the pattern simply omits the field.
-        try
-        {
-            AutomationElement doc = root.FindFirst(TreeScope.Descendants,
-                new PropertyCondition(AutomationElement.ControlTypeProperty, ControlType.Document));
-            if (doc != null)
-            {
-                object dp;
-                if (doc.TryGetCurrentPattern(TextPattern.Pattern, out dp))
-                {
-                    string body = ((TextPattern)dp).DocumentRange.GetText(8000);
-                    if (!string.IsNullOrEmpty(body)) state["document_text"] = body;
-                }
-            }
-            object focusPattern;
-            AutomationElement focusedElement = AutomationElement.FocusedElement;
-            if (focusedElement != null && focusedElement.TryGetCurrentPattern(TextPattern.Pattern, out focusPattern))
-            {
-                System.Windows.Automation.Text.TextPatternRange[] ranges = ((TextPattern)focusPattern).GetSelection();
-                if (ranges != null && ranges.Length > 0)
-                {
-                    string selected = ranges[0].GetText(4000);
-                    if (!string.IsNullOrEmpty(selected)) state["selected_text"] = selected;
-                }
-            }
-            Condition selectable = new PropertyCondition(AutomationElement.IsSelectionItemPatternAvailableProperty, true);
-            AutomationElementCollection selectedItems = root.FindAll(TreeScope.Descendants, selectable);
-            List<object> selectedLines = new List<object>();
-            for (int i = 0; i < selectedItems.Count && selectedLines.Count < 20; i++)
-            {
-                try
-                {
-                    object sp;
-                    if (selectedItems[i].TryGetCurrentPattern(SelectionItemPattern.Pattern, out sp) && ((SelectionItemPattern)sp).Current.IsSelected)
-                        selectedLines.Add(SafeString(selectedItems[i].Current.Name) + " (" + SafeRole(selectedItems[i].Current) + ")");
-                }
-                catch { }
-            }
-            if (selectedLines.Count > 0) state["selected_elements"] = selectedLines;
-        }
-        catch { }
-        state["snapshotAt"] = snap.At.ToString("o");
-        state["ttlMs"] = (long)ttlMs;
-        if (focused != null) state["focused_element"] = focused;
-        return state;
-    }
-
-    void Walk(AutomationElement el, int depth, long maxElements, Snapshot snap, StringBuilder sb, ref int index, IntPtr hwnd, ref string focused)
-    {
-        if (el == null) return;
-        AutomationElementCollection kids;
-        try { kids = el.FindAll(TreeScope.Children, Condition.TrueCondition); }
-        catch { return; }
-
-        for (int i = 0; i < kids.Count; i++)
-        {
-            if (index >= maxElements) return;
-            AutomationElement k;
-            try { k = kids[i]; } catch { continue; }
-            AutomationElement.AutomationElementInformation info;
-            bool offscreen;
-            try { info = k.Current; offscreen = info.IsOffscreen; } catch { continue; }
-            if (offscreen) continue;
-            try
-            {
-
-            long idx = index++;
-            snap.Elements[idx] = k;
-
-            string role = SafeRole(info);
-            string name = SafeString(info.Name);
-            System.Windows.Rect rect = System.Windows.Rect.Empty;
-            try { rect = info.BoundingRectangle; } catch { }
-            bool hasRect = !rect.IsEmpty && rect.Width > 0 && rect.Height > 0;
-
-            Dictionary<string, object> summary = new Dictionary<string, object>();
-            summary["index"] = idx;
-            summary["role"] = role;
-            summary["name"] = name;
-            if (hasRect)
-            {
-                summary["x"] = (int)rect.X;
-                summary["y"] = (int)rect.Y;
-                summary["width"] = (int)rect.Width;
-                summary["height"] = (int)rect.Height;
-            }
-            snap.Summary.Add(summary);
-
-            sb.Append('[').Append(idx).Append("] ").Append(role);
-            if (name.Length > 0) sb.Append(" \"").Append(Trim(name, 120)).Append('"');
-            if (hasRect) sb.Append(" @").Append((int)rect.X).Append(',').Append((int)rect.Y).Append(' ').Append((int)rect.Width).Append('x').Append((int)rect.Height);
-
-            List<string> actions = ActionsOf(k);
-            if (actions.Count > 0) sb.Append(" [").Append(string.Join(",", actions.ToArray())).Append(']');
-
-            StringBuilder val = new StringBuilder();
-            try
-            {
-                object vp;
-                if (k.TryGetCurrentPattern(ValuePattern.Pattern, out vp))
-                {
-                    string v = ((ValuePattern)vp).Current.Value;
-                    if (!string.IsNullOrEmpty(v)) val.Append(" value=\"").Append(Trim(v, 90)).Append('"');
-                }
-                else
-                {
-                    object tp;
-                    if (k.TryGetCurrentPattern(TextPattern.Pattern, out tp))
-                    {
-                        string t = ((TextPattern)tp).DocumentRange.GetText(120);
-                        if (!string.IsNullOrEmpty(t)) val.Append(" text=\"").Append(Trim(t.Replace("\r", " ").Replace("\n", " "), 90)).Append('"');
-                    }
-                }
-            }
-            catch { }
-            sb.Append(val.ToString());
-
-            try { if (info.HasKeyboardFocus && focused == null) focused = "[" + idx + "] " + role + " \"" + name + "\""; }
-            catch { }
-
-            sb.Append('\n');
-            if (depth < 18) Walk(k, depth + 1, maxElements, snap, sb, ref index, hwnd, ref focused);
-            }
-            catch (Exception) { continue; }
-        }
-    }
-
-    static string SafeRole(AutomationElement.AutomationElementInformation info)
-    {
-        try { return info.ControlType.ProgrammaticName.Replace("ControlType.", ""); }
-        catch { return "Unknown"; }
-    }
-
-    static long ToLong(object v) { try { return Convert.ToInt64(v, System.Globalization.CultureInfo.InvariantCulture); } catch { return 0L; } }
-    static double ToDouble(object v) { try { return Convert.ToDouble(v, System.Globalization.CultureInfo.InvariantCulture); } catch { return 1.0; } }
-
-    static string SafeString(string s) { return s == null ? "" : s.Replace("\r", " ").Replace("\n", " ").Trim(); }
-
-    static string Trim(string s, int max) { return s.Length <= max ? s : s.Substring(0, max) + "…"; }
-
-    static List<string> ActionsOf(AutomationElement el)
-    {
-        List<string> a = new List<string>();
-        object pattern;
-        try { if (el.TryGetCurrentPattern(InvokePattern.Pattern, out pattern)) a.Add("Invoke"); } catch { }
-        try { if (el.TryGetCurrentPattern(TogglePattern.Pattern, out pattern)) a.Add("Toggle"); } catch { }
-        try { if (el.TryGetCurrentPattern(SelectionItemPattern.Pattern, out pattern)) a.Add("Select"); } catch { }
-        try { if (el.TryGetCurrentPattern(ExpandCollapsePattern.Pattern, out pattern)) a.Add("ExpandCollapse"); } catch { }
-        try { if (el.TryGetCurrentPattern(ScrollPattern.Pattern, out pattern)) a.Add("Scroll"); } catch { }
-        try { if (el.TryGetCurrentPattern(ValuePattern.Pattern, out pattern)) a.Add("SetValue"); } catch { }
-        return a;
-    }
-
-    /// Every top-level window that belongs to the same "visual state" as the
-    /// target: the window itself plus owned/transient popups (menus, tooltips,
-    /// combo drop-downs) that a screenshot of the window alone would hide.
-    List<IntPtr> RelatedWindows(IntPtr target)
-    {
-        RECT main = BoundsOf(target);
-        uint targetPid;
-        GetWindowThreadProcessId(target, out targetPid);
-        List<IntPtr> result = new List<IntPtr>();
-        List<IntPtr> all = EnumerateWindows(false);
-        for (int i = 0; i < all.Count; i++)
-        {
-            IntPtr h = all[i];
-            if (h == target) continue;
-            if (!IsWindowVisible(h)) continue;
-            if (IsIconic(h)) continue;
-            uint pid;
-            GetWindowThreadProcessId(h, out pid);
-            if (pid != targetPid) continue;
-            bool owned = GetWindow(h, GW_OWNER) == target;
-            bool popup = (GetWindowLong(h, GWL_STYLE) & WS_POPUP) != 0;
-            if (!owned && !popup) continue;
-            RECT r = BoundsOf(h);
-            bool intersects = r.Left < main.Right && r.Right > main.Left && r.Top < main.Bottom && r.Bottom > main.Top;
-            if (!intersects) continue;
-            result.Add(h);
-        }
-        return result;
-    }
-
-    List<object> CaptureScreenshots(IntPtr h, out bool anyBlank)
-    {
-        anyBlank = false;
-        if (IsIconic(h))
-            throw new ToolError("window_minimized", "window " + h.ToInt64() + " is minimized, so there is nothing to observe: call activate_window (or restore it) and observe again. A minimized window must never be captured, because the image would not be the window.");
-        if (!IsWindowVisible(h))
-            throw new ToolError("window_not_visible", "window " + h.ToInt64() + " is not visible; activate it before observing.");
-
-        List<IntPtr> order = EnumerateWindows(true); // z-order, topmost first
-        List<object> shots = new List<object>();
-        List<IntPtr> targets = new List<IntPtr>();
-        targets.Add(h);
-        targets.AddRange(RelatedWindows(h));
-
-        int total = Math.Max(1, order.Count);
-        for (int i = 0; i < targets.Count; i++)
-        {
-            IntPtr w = targets[i];
-            int zIndex = order.IndexOf(w);
-            if (zIndex < 0) zIndex = total - 1;
-            Dictionary<string, object> shot = CaptureWindow(w);
-            shot["zIndex"] = (long)(total - zIndex); // larger = visually above
-            string id = "shot-" + _shotGeneration + "-" + (++_shotSeq).ToString() + "-" + w.ToInt64().ToString();
-            shot["id"] = id;
-            RECT r = BoundsOf(w);
-            lock (_sync)
-            {
-                PruneShots();
-                _shots[id] = new ShotRegion
-                {
-                    WindowId = h.ToInt64(), OriginX = r.Left, OriginY = r.Top,
-                    Width = (int)ToLong(shot["width"]), Height = (int)ToLong(shot["height"]),
-                    ScaleX = ToDouble(shot["scaleX"]), ScaleY = ToDouble(shot["scaleY"]),
-                    At = DateTime.UtcNow,
-                };
-            }
-            shots.Add(shot);
-        }
-        return shots;
-    }
-
-    void PruneShots()
-    {
-        if (_shots.Count <= 48) return;
-        List<string> stale = new List<string>();
-        DateTime cutoff = DateTime.UtcNow.AddMilliseconds(-_shotTtlMs);
-        foreach (KeyValuePair<string, ShotRegion> kv in _shots)
-            if (kv.Value.At < cutoff) stale.Add(kv.Key);
-        for (int i = 0; i < stale.Count; i++) _shots.Remove(stale[i]);
-    }
-
-    ShotRegion ResolveShot(IntPtr h, string screenshotId)
-    {
-        if (string.IsNullOrEmpty(screenshotId))
-            throw new ToolError("bad_request", "screenshotId is empty");
-        ShotRegion region;
-        lock (_sync)
-        {
-            if (!_shots.TryGetValue(screenshotId, out region))
-                throw new ToolError("stale_screenshot", "screenshot '" + screenshotId + "' is no longer cached: it must come from the latest get_window_state for this window. Take a fresh screenshot before retrying.");
-        }
-        if (region.WindowId != h.ToInt64())
-            throw new ToolError("screenshot_mismatch", "screenshot '" + screenshotId + "' belongs to another window; take a fresh observation for this window.");
-        if ((DateTime.UtcNow - region.At).TotalMilliseconds > _shotTtlMs)
-            throw new ToolError("stale_screenshot", "screenshot '" + screenshotId + "' is older than the screenshot TTL. Take a fresh screenshot before retrying.");
-        return region;
-    }
-
-    /// Codex rejects a point that falls outside the screenshot the caller referenced
-    /// ("is outside screenshot bounds"), because that usually means a stale observation.
-    static void AssertPointInsideShot(double x, double y, ShotRegion region, string screenshotId, string what)
-    {
-        if (double.IsNaN(x) || double.IsNaN(y)) return;
-        if (x >= 0 && y >= 0 && x <= region.Width && y <= region.Height) return;
-        throw new ToolError("point_outside_screenshot",
-            what + " (" + x.ToString("0.##") + ", " + y.ToString("0.##") + ") is outside screenshot bounds (0,0)-(" +
-            region.Width + "," + region.Height + ") for '" + screenshotId + "'. Take a fresh screenshot with get_window_state before retrying.");
-    }
-
-    /// Windows.Graphics.Capture through the .NET 8 helper: composed window content,
-    /// occlusion-proof, no DPI halo. Falls back to PrintWindow/screen when absent.
-    static Bitmap TryCaptureWithWgc(IntPtr hwnd)
-    {
-        try
-        {
-            string dir = Path.GetDirectoryName(typeof(WindowsDriver).Assembly.Location) ?? ".";
-            string exe = Path.Combine(dir, "capture", "dsh-cu-capture.exe");
-            if (!File.Exists(exe)) return null;
-            string tmp = Path.Combine(Path.GetTempPath(), "dsh-cu-wgc-" + Guid.NewGuid().ToString("N") + ".png");
-            try
-            {
-                long budget = Budget.RemainingMs();
-                int waitMs = (int)Math.Max(150, Math.Min(1200, budget - 150));
-                ProcessStartInfo psi = new ProcessStartInfo(exe, "--hwnd " + hwnd.ToInt64().ToString() + " --out \"" + tmp + "\" --timeout " + waitMs);
-                psi.UseShellExecute = false;
-                psi.CreateNoWindow = true;
-                psi.RedirectStandardOutput = true;
-                psi.RedirectStandardError = true;
-                using (Process p = Process.Start(psi))
-                {
-                    if (p == null) return null;
-                    if (!p.WaitForExit(waitMs + 900)) { try { p.Kill(); } catch { } return null; }
-                    if (p.ExitCode != 0 || !File.Exists(tmp)) return null;
-                    byte[] bytes = File.ReadAllBytes(tmp);
-                    using (MemoryStream ms = new MemoryStream(bytes)) return new Bitmap(ms);
-                }
-            }
-            finally { try { if (File.Exists(tmp)) File.Delete(tmp); } catch { } }
-        }
-        catch { return null; }
-    }
-
-    Dictionary<string, object> CaptureWindow(IntPtr h)
-    {
-        RECT r = BoundsOf(h);
-        int w = Math.Max(1, r.Right - r.Left);
-        int ht = Math.Max(1, r.Bottom - r.Top);
-        // Cap pathological sizes so a maximised multi-monitor window cannot blow the
-        // attachment byte budget.
-        const int cap = 2560;
-        double scale = 1.0;
-        if (w > cap || ht > cap)
-        {
-            scale = Math.Min((double)cap / w, (double)cap / ht);
-            w = Math.Max(1, (int)(w * scale));
-            ht = Math.Max(1, (int)(ht * scale));
-        }
-
-        Bitmap wgc = TryCaptureWithWgc(h);
-        if (wgc != null)
-        {
-            Dictionary<string, object> wg = FinishShot(h, wgc, r, true, 1.0, 1.0);
-            return wg;
-        }
-
-        Bitmap bmp = new Bitmap(w, ht, PixelFormat.Format24bppRgb);
-        bool printed = false;
-        using (Graphics g = Graphics.FromImage(bmp))
-        {
-            IntPtr hdc = g.GetHdc();
-            try { printed = PrintWindow(h, hdc, 2); }
-            finally { g.ReleaseHdc(hdc); }
-        }
-        if (!printed || IsBlank(bmp))
-        {
-            bmp.Dispose();
-            bmp = CaptureFromScreen(r, w, ht);
-        }
-
-        // DPI-virtualized targets render at their own (usually 96 DPI) size while
-        // occupying a larger physical rect, so PrintWindow leaves a black halo.
-        // Crop to the painted region and remember the physical/screenshot ratio so
-        // coordinate actions can map screenshot pixels back to screen pixels.
-        int fullWidth = bmp.Width, fullHeight = bmp.Height;
-        Rectangle painted = PaintedBounds(bmp);
-        if (painted.Width > 16 && painted.Height > 16 && (painted.Width < bmp.Width || painted.Height < bmp.Height))
-        {
-            Bitmap cropped = bmp.Clone(new Rectangle(0, 0, painted.Right, painted.Bottom), bmp.PixelFormat);
-            bmp.Dispose();
-            bmp = cropped;
-            w = bmp.Width;
-            ht = bmp.Height;
-        }
-        double scaleX = (double)fullWidth / Math.Max(1, bmp.Width);
-        double scaleY = (double)fullHeight / Math.Max(1, bmp.Height);
-
-        byte[] png;
-        using (MemoryStream ms = new MemoryStream())
-        {
-            bmp.Save(ms, ImageFormat.Png);
-            png = ms.ToArray();
-        }
-        bmp.Dispose();
-
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["mediaType"] = "image/png";
-        d["base64"] = Convert.ToBase64String(png);
-        d["width"] = (long)w;
-        d["height"] = (long)ht;
-        d["originX"] = (long)r.Left;
-        d["originY"] = (long)r.Top;
-        d["zIndex"] = 1L;
-        d["printed"] = printed;
-        d["scaleX"] = scaleX;
-        d["scaleY"] = scaleY;
-        return d;
-    }
-
-    static Dictionary<string, object> FinishShot(IntPtr h, Bitmap bmp, RECT r, bool printed, double scaleX, double scaleY)
-    {
-        int w = bmp.Width, ht = bmp.Height;
-        byte[] png;
-        using (MemoryStream ms = new MemoryStream())
-        {
-            bmp.Save(ms, ImageFormat.Png);
-            png = ms.ToArray();
-        }
-        bmp.Dispose();
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["mediaType"] = "image/png";
-        d["base64"] = Convert.ToBase64String(png);
-        d["width"] = (long)w;
-        d["height"] = (long)ht;
-        d["originX"] = (long)r.Left;
-        d["originY"] = (long)r.Top;
-        d["zIndex"] = 1L;
-        d["printed"] = printed;
-        d["scaleX"] = scaleX;
-        d["scaleY"] = scaleY;
-        return d;
-    }
-
-    static Bitmap CaptureFromScreen(RECT r, int w, int ht)
-    {
-        Bitmap bmp = new Bitmap(w, ht, PixelFormat.Format24bppRgb);
-        using (Graphics g = Graphics.FromImage(bmp))
-        {
-            try { g.CopyFromScreen(r.Left, r.Top, 0, 0, new Size(w, ht), CopyPixelOperation.SourceCopy); }
-            catch { }
-        }
-        return bmp;
-    }
-
-    /// Bounding box of everything that is not uniform black, i.e. the region the
-    /// target actually painted. Used to strip the DPI-virtualization halo.
-    static Rectangle PaintedBounds(Bitmap bmp)
-    {
-        int right = 0, bottom = 0;
-        int step = Math.Max(1, Math.Min(bmp.Width, bmp.Height) / 200);
-        for (int y = 0; y < bmp.Height; y += step)
-        {
-            for (int x = 0; x < bmp.Width; x += step)
-            {
-                Color c = bmp.GetPixel(x, y);
-                if (c.R > 8 || c.G > 8 || c.B > 8)
-                {
-                    if (x > right) right = x;
-                    if (y > bottom) bottom = y;
-                }
-            }
-        }
-        return new Rectangle(0, 0, Math.Min(bmp.Width, right + step), Math.Min(bmp.Height, bottom + step));
-    }
-
-    static bool IsBlank(Bitmap bmp)
-    {
-        int samples = 0, dark = 0;
-        int stepX = Math.Max(1, bmp.Width / 12);
-        int stepY = Math.Max(1, bmp.Height / 12);
-        for (int y = 0; y < bmp.Height; y += stepY)
-        {
-            for (int x = 0; x < bmp.Width; x += stepX)
-            {
-                Color c = bmp.GetPixel(x, y);
-                samples++;
-                if (c.R < 6 && c.G < 6 && c.B < 6) dark++;
-            }
-        }
-        return samples > 0 && dark == samples;
-    }
-
-    // ---- element resolution ----
-    AutomationElement ElementAt(IntPtr hwnd, long index)
-    {
-        Snapshot snap;
-        lock (_sync)
-        {
-            if (!_snapshots.TryGetValue(hwnd.ToInt64(), out snap))
-                throw new ToolError("stale_snapshot", "no observation for this window yet: call get_window_state with include_text=true before using element_index");
-        }
-        if ((DateTime.UtcNow - snap.At).TotalMilliseconds > _snapshotTtlMs)
-            throw new ToolError("stale_snapshot", "the element snapshot is older than " + _snapshotTtlMs + "ms: call get_window_state again before acting on element indexes");
-        AutomationElement el;
-        if (!snap.Elements.TryGetValue(index, out el))
-            throw new ToolError("bad_index", "element index " + index + " is not part of the latest observation for this window");
-        return el;
-    }
-
-    static Dictionary<string, object> ScreenshotTarget(IntPtr h, AutomationElement el)
-    {
-        RECT obs = BoundsOf(h);
-        System.Windows.Rect r = el.Current.BoundingRectangle;
-        if (r.IsEmpty || r.Width <= 0 || r.Height <= 0)
-            throw new ToolError("no_geometry", "the element has no bounding rectangle; use x/y coordinates instead");
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["screenX"] = (int)r.X + (int)(r.Width / 2);
-        d["screenY"] = (int)r.Y + (int)(r.Height / 2);
-        d["originX"] = obs.Left;
-        d["originY"] = obs.Top;
-        return d;
-    }
-
-    // ---- input ----
-    static void SendMouse(uint flags) { SendMouseData(flags, 0); }
-
-    static void SendMouseData(uint flags, int data)
-    {
-        INPUT inp = new INPUT();
-        inp.type = INPUT_MOUSE;
-        inp.u.mi.dx = 0; inp.u.mi.dy = 0;
-        inp.u.mi.mouseData = (uint)data;
-        inp.u.mi.dwFlags = flags;
-        INPUT[] arr = new INPUT[] { inp };
-        SendInput(1, arr, Marshal.SizeOf(typeof(INPUT)));
-    }
-
-    static void KeyDown(ushort vk) { SendKey(vk, 0, 0); }
-    static void KeyUp(ushort vk) { SendKey(vk, 0, KEYEVENTF_KEYUP); }
-
-    static void SendKey(ushort vk, ushort scan, uint flags)
-    {
-        INPUT inp = new INPUT();
-        inp.type = INPUT_KEYBOARD;
-        inp.u.ki.wVk = vk;
-        inp.u.ki.wScan = scan;
-        inp.u.ki.dwFlags = flags;
-        INPUT[] arr = new INPUT[] { inp };
-        SendInput(1, arr, Marshal.SizeOf(typeof(INPUT)));
-    }
-
-    static void SendUnicode(char ch, bool up)
-    {
-        INPUT inp = new INPUT();
-        inp.type = INPUT_KEYBOARD;
-        inp.u.ki.wVk = 0;
-        inp.u.ki.wScan = ch;
-        inp.u.ki.dwFlags = KEYEVENTF_UNICODE | (up ? KEYEVENTF_KEYUP : 0);
-        INPUT[] arr = new INPUT[] { inp };
-        SendInput(1, arr, Marshal.SizeOf(typeof(INPUT)));
-    }
-
-    void ClickAt(int screenX, int screenY, int count, string button, Dictionary<string, object> p)
-    {
-        SetCursorPos(screenX, screenY);
-        System.Threading.Thread.Sleep(35);
-        for (int i = 0; i < Math.Max(1, count); i++)
-        {
-            CheckCancel(p);
-            if (i > 0) System.Threading.Thread.Sleep(60);
-            uint down = MOUSEEVENTF_LEFTDOWN, up = MOUSEEVENTF_LEFTUP;
-            if (button == "r") { down = MOUSEEVENTF_RIGHTDOWN; up = MOUSEEVENTF_RIGHTUP; }
-            else if (button == "m") { down = MOUSEEVENTF_MIDDLEDOWN; up = MOUSEEVENTF_MIDDLEUP; }
-            SendMouse(down);
-            try { System.Threading.Thread.Sleep(22); }
-            finally { SendMouse(up); }
-        }
-    }
-
-    public Dictionary<string, object> Click(Dictionary<string, object> p)
-    {
-        CheckCancel(p);
-        AssertUnlocked();
-        EnsureOverlay(8000);
-        IntPtr h = ResolveWindow(p);
-        AssertAppApproved(h);
-        long elementIndex = Json.GetLong(p, "element_index", -1);
-        string button = NormalizeButton(Json.GetString(p, "mouse_button"));
-        int count = (int)Json.GetLong(p, "click_count", 1);
-        int sx, sy;
-
-        if (elementIndex >= 0)
-        {
-            AutomationElement el = ElementAt(h, elementIndex);
-            Dictionary<string, object> t = ScreenshotTarget(h, el);
-            sx = (int)t["screenX"]; sy = (int)t["screenY"];
-            // Prefer the semantic invoke when the element exposes one and we are
-            // doing a plain left single click: it survives occlusion and does not
-            // depend on pixel geometry.
-            if (count == 1 && button == "l")
-            {
-                object pattern;
-                try
-                {
-                    if (el.TryGetCurrentPattern(InvokePattern.Pattern, out pattern))
-                    {
-                        ((InvokePattern)pattern).Invoke();
-                        Dictionary<string, object> r2 = new Dictionary<string, object>();
-                        r2["clicked"] = "invoke"; r2["element_index"] = elementIndex;
-                        return r2;
-                    }
-                }
-                catch (Exception) { }
-            }
-        }
-        else
-        {
-            double x = Json.GetDouble(p, "x", double.NaN);
-            double y = Json.GetDouble(p, "y", double.NaN);
-            if (double.IsNaN(x) || double.IsNaN(y)) throw new ToolError("bad_request", "click needs either 'element_index' or both 'x' and 'y' (window-relative pixels)");
-            string screenshotId = Json.GetString(p, "screenshotId");
-            int originX, originY;
-            double coordScaleX = 1.0, coordScaleY = 1.0;
-            if (!string.IsNullOrEmpty(screenshotId))
-            {
-                ShotRegion region = ResolveShot(h, screenshotId);
-                AssertPointInsideShot(x, y, region, screenshotId, "click");
-                originX = region.OriginX; originY = region.OriginY;
-                coordScaleX = region.ScaleX; coordScaleY = region.ScaleY;
-            }
-            else
-            {
-                RECT bounds = BoundsOf(h);
-                originX = bounds.Left; originY = bounds.Top;
-            }
-            sx = originX + (int)Math.Round(x * coordScaleX);
-            sy = originY + (int)Math.Round(y * coordScaleY);
-        }
-
-        Activate(h); // pointer clicks need a real foreground window
-        MarkCursor(sx, sy);
-        CheckCancel(p);
-        ClickAt(sx, sy, count, button, p);
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["clicked"] = "pointer";
-        r["screenX"] = (long)sx;
-        r["screenY"] = (long)sy;
-        r["clickCount"] = (long)count;
-        return r;
-    }
-
-    static string NormalizeButton(string b)
-    {
-        if (string.IsNullOrEmpty(b)) return "l";
-        string s = b.Trim().ToLowerInvariant();
-        if (s == "right" || s == "r") return "r";
-        if (s == "middle" || s == "m") return "m";
-        return "l";
-    }
-
-    public Dictionary<string, object> PressKey(Dictionary<string, object> p)
-    {
-        CheckCancel(p);
-        AssertUnlocked();
-        EnsureOverlay(8000);
-        IntPtr h = ResolveWindow(p);
-        AssertAppApproved(h);
-        string key = Json.GetString(p, "key");
-        if (string.IsNullOrEmpty(key)) throw new ToolError("bad_request", "press_key requires 'key'");
-        Activate(h);
-        Chord(key, p);
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["pressed"] = key;
-        return r;
-    }
-
-    static readonly Dictionary<string, ushort> KeyMap = BuildKeyMap();
-
-    static Dictionary<string, ushort> BuildKeyMap()
-    {
-        Dictionary<string, ushort> m = new Dictionary<string, ushort>(StringComparer.OrdinalIgnoreCase);
-        for (char c = 'a'; c <= 'z'; c++) m[c.ToString()] = (ushort)(0x41 + (c - 'a'));
-        for (char c = 'A'; c <= 'Z'; c++) m[c.ToString()] = (ushort)(0x41 + (c - 'A'));
-        for (char c = '0'; c <= '9'; c++) m[c.ToString()] = (ushort)(0x30 + (c - '0'));
-        m["return"] = 0x0D; m["enter"] = 0x0D; m["kp_enter"] = 0x0D;
-        m["tab"] = 0x09; m["space"] = 0x20; m["backspace"] = 0x08; m["escape"] = 0x1B; m["esc"] = 0x1B;
-        m["delete"] = 0x2E; m["insert"] = 0x2D; m["home"] = 0x24; m["end"] = 0x23;
-        m["page_up"] = 0x21; m["pageup"] = 0x21; m["prior"] = 0x21; m["page_down"] = 0x22; m["pagedown"] = 0x22; m["next"] = 0x22;
-        m["left"] = 0x25; m["up"] = 0x26; m["right"] = 0x27; m["down"] = 0x28;
-        m["shift_l"] = 0xA0; m["shift"] = 0xA0; m["shift_r"] = 0xA1;
-        m["control_l"] = 0xA2; m["control"] = 0xA2; m["ctrl"] = 0xA2; m["control_r"] = 0xA3;
-        m["alt_l"] = 0xA4; m["alt"] = 0xA4; m["alt_r"] = 0xA5;
-        m["super_l"] = 0x5B; m["win"] = 0x5B; m["meta"] = 0x5B; m["lwin"] = 0x5B; m["super"] = 0x5B;
-        m["super_r"] = 0x5C; m["rwin"] = 0x5C;
-        m["caps_lock"] = 0x14; m["num_lock"] = 0x90; m["scroll_lock"] = 0x91;
-        m["menu"] = 0x5D; m["apps"] = 0x5D; m["print"] = 0x2C; m["pause"] = 0x13;
-        m["period"] = 0xBE; m["greater"] = 0xBE; m["comma"] = 0xBC; m["less"] = 0xBC;
-        m["slash"] = 0xBF; m["question"] = 0xBF; m["backslash"] = 0xDC; m["bar"] = 0xDC;
-        m["semicolon"] = 0xBA; m["colon"] = 0xBA; m["apostrophe"] = 0xDE; m["quote"] = 0xDE;
-        m["minus"] = 0xBD; m["underscore"] = 0xBD; m["equal"] = 0xBB; m["plus"] = 0xBB;
-        m["bracketleft"] = 0xDB; m["bracketright"] = 0xDD; m["grave"] = 0xC0; m["asciitilde"] = 0xC0;
-        for (int i = 1; i <= 24; i++) { m["f" + i] = (ushort)(0x6F + i); }
-        // Numpad: both X keysym style (KP_0) and the aliases the API documents.
-        for (int i = 0; i <= 9; i++)
-        {
-            m["kp_" + i] = (ushort)(0x60 + i);
-            m["numpad_" + i] = (ushort)(0x60 + i);
-        }
-        m["kp_add"] = 0x6B; m["numpad_add"] = 0x6B; m["plus_numpad"] = 0x6B;
-        m["kp_subtract"] = 0x6D; m["numpad_subtract"] = 0x6D; m["minus_numpad"] = 0x6D;
-        m["kp_multiply"] = 0x6A; m["numpad_multiply"] = 0x6A; m["kp_asterisk"] = 0x6A;
-        m["kp_divide"] = 0x6F; m["numpad_divide"] = 0x6F; m["kp_slash"] = 0x6F;
-        m["kp_decimal"] = 0x6E; m["numpad_decimal"] = 0x6E; m["kp_separator"] = 0x6C;
-        m["numpad_enter"] = 0x0D;
-        return m;
-    }
-
-    static readonly HashSet<ushort> ExtendedKeys = new HashSet<ushort>(new ushort[] {
-        0x21,0x22,0x23,0x24,0x25,0x26,0x27,0x28,0x2D,0x2E,0x5B,0x5C,0x5D,0x90,0x6F,0xA3,0xA5
-    });
-
-    static void SendVk(ushort vk, bool up)
-    {
-        uint flags = up ? KEYEVENTF_KEYUP : 0;
-        if (ExtendedKeys.Contains(vk)) flags |= KEYEVENTF_EXTENDEDKEY;
-        SendKey(vk, 0, flags);
-    }
-
-    void Chord(string chord, Dictionary<string, object> p)
-    {
-        string[] parts = chord.Split('+');
-        List<ushort> keys = new List<ushort>();
-        foreach (string raw in parts)
-        {
-            string token = raw.Trim();
-            if (token.Length == 0) continue;
-            keys.Add(ResolveKey(token));
-        }
-        if (keys.Count == 0) throw new ToolError("bad_request", "empty key chord");
-        int pressed = 0;
-        try
-        {
-            for (int i = 0; i < keys.Count; i++)
-            {
-                CheckCancel(p);
-                SendVk(keys[i], false); pressed++;
-                System.Threading.Thread.Sleep(12);
-            }
-        }
-        finally { for (int i = pressed - 1; i >= 0; i--) SendVk(keys[i], true); }
-    }
-
-    /// The Windows/Meta key family is hard-denied: it opens Run, Settings and
-    /// shell surfaces that Computer Use must never drive. Enforced in the helper
-    /// as well as in the plugin, so a direct protocol call cannot bypass it.
-    static readonly HashSet<string> DeniedKeys = new HashSet<string>(StringComparer.OrdinalIgnoreCase) {
-        "win", "win_l", "win_r", "lwin", "rwin", "windows", "windows_l", "windows_r",
-        "super", "super_l", "super_r", "meta", "meta_l", "meta_r", "cmd", "command", "os", "os_l", "os_r"
-    };
-
-    static ushort ResolveKey(string token)
-    {
-        if (DeniedKeys.Contains(token))
-            throw new ToolError("hard_denied", "the Windows/Meta/Super key ('" + token + "') is hard-denied: it opens Run, Settings and shell surfaces that Computer Use must never drive.");
-        ushort vk;
-        if (KeyMap.TryGetValue(token, out vk)) return vk;
-        if (token.Length == 1)
-        {
-            char c = token[0];
-            short scan = VkKeyScan(c);
-            if (scan != -1) return (ushort)(scan & 0xFF);
-            return (ushort)char.ToUpperInvariant(c);
-        }
-        throw new ToolError("bad_key", "unknown key name '" + token + "'");
-    }
-
-    [DllImport("user32.dll")] static extern short VkKeyScan(char ch);
-
-    public Dictionary<string, object> TypeText(Dictionary<string, object> p)
-    {
-        CheckCancel(p);
-        AssertUnlocked();
-        EnsureOverlay(8000);
-        IntPtr h = ResolveWindow(p);
-        AssertAppApproved(h);
-        string text = Json.GetString(p, "text");
-        if (text == null) throw new ToolError("bad_request", "type_text requires 'text'");
-        Activate(h);
-        for (int i = 0; i < text.Length; i++)
-        {
-            CheckCancel(p);
-            char c = text[i];
-            if (c == '\n') { SendVk(0x0D, false); SendVk(0x0D, true); continue; }
-            if (c == '\t') { SendVk(0x09, false); SendVk(0x09, true); continue; }
-            if (c == '\r') continue;
-            SendUnicode(c, false);
-            SendUnicode(c, true);
-            if ((i & 15) == 0) System.Threading.Thread.Sleep(4);
-        }
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["typed"] = (long)text.Length;
-        return r;
-    }
-
-    public Dictionary<string, object> Scroll(Dictionary<string, object> p)
-    {
-        CheckCancel(p);
-        AssertUnlocked();
-        EnsureOverlay(8000);
-        IntPtr h = ResolveWindow(p);
-        AssertAppApproved(h);
-        // Validation first: an invalid call must not activate a window or move the pointer.
-        double scrollY = Json.GetDouble(p, "scrollY", 0);
-        double scrollX = Json.GetDouble(p, "scrollX", 0);
-        double x = Json.GetDouble(p, "x", double.NaN);
-        double y = Json.GetDouble(p, "y", double.NaN);
-        if (double.IsNaN(x) || double.IsNaN(y))
-            throw new ToolError("bad_request", "scroll requires finite 'x' and 'y' (the point to scroll from, in window/screenshot coordinates); click a pane first if it needs focus");
-        if (Math.Abs(scrollX) < 0.0001 && Math.Abs(scrollY) < 0.0001)
-            throw new ToolError("bad_request", "scroll requires a non-zero scrollX or scrollY");
-        Activate(h);
-        int sx, sy;
-        string scrollShot = Json.GetString(p, "screenshotId");
-        int originLeft, originTop;
-        double scrollScaleX = 1.0, scrollScaleY = 1.0;
-        if (!string.IsNullOrEmpty(scrollShot))
-        {
-            ShotRegion region = ResolveShot(h, scrollShot);
-            AssertPointInsideShot(x, y, region, scrollShot, "scroll origin");
-            originLeft = region.OriginX; originTop = region.OriginY;
-            scrollScaleX = region.ScaleX; scrollScaleY = region.ScaleY;
-        }
-        else
-        {
-            RECT b = BoundsOf(h);
-            originLeft = b.Left; originTop = b.Top;
-        }
-        sx = originLeft + (int)Math.Round(x * scrollScaleX);
-        sy = originTop + (int)Math.Round(y * scrollScaleY);
-        MarkCursor(sx, sy);
-        SetCursorPos(sx, sy);
-        System.Threading.Thread.Sleep(40);
-        if (Math.Abs(scrollY) > 0.0001) SendMouseData(MOUSEEVENTF_WHEEL, (int)(-scrollY * 120));
-        if (Math.Abs(scrollX) > 0.0001) SendMouseData(MOUSEEVENTF_HWHEEL, (int)(-scrollX * 120));
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["scrolled"] = true;
-        r["screenX"] = (long)sx; r["screenY"] = (long)sy;
-        return r;
-    }
-
-    public Dictionary<string, object> Drag(Dictionary<string, object> p)
-    {
-        CheckCancel(p);
-        AssertUnlocked();
-        EnsureOverlay(8000);
-        IntPtr h = ResolveWindow(p);
-        AssertAppApproved(h);
-        double fromX = Json.GetDouble(p, "from_x", double.NaN);
-        double fromY = Json.GetDouble(p, "from_y", double.NaN);
-        double toX = Json.GetDouble(p, "to_x", double.NaN);
-        double toY = Json.GetDouble(p, "to_y", double.NaN);
-        if (double.IsNaN(fromX) || double.IsNaN(fromY) || double.IsNaN(toX) || double.IsNaN(toY))
-            throw new ToolError("bad_request", "drag requires finite from_x/from_y/to_x/to_y (window-relative pixels)");
-        Activate(h);
-        int originX, originY;
-        double dragScaleX = 1.0, dragScaleY = 1.0;
-        string dragShot = Json.GetString(p, "screenshotId");
-        if (!string.IsNullOrEmpty(dragShot))
-        {
-            ShotRegion region = ResolveShot(h, dragShot);
-            AssertPointInsideShot(fromX, fromY, region, dragShot, "drag start");
-            AssertPointInsideShot(toX, toY, region, dragShot, "drag end");
-            originX = region.OriginX; originY = region.OriginY;
-            dragScaleX = region.ScaleX; dragScaleY = region.ScaleY;
-        }
-        else
-        {
-            RECT b = BoundsOf(h);
-            originX = b.Left; originY = b.Top;
-        }
-        int fx = originX + (int)Math.Round(fromX * dragScaleX);
-        int fy = originY + (int)Math.Round(fromY * dragScaleY);
-        int tx = originX + (int)Math.Round(toX * dragScaleX);
-        int ty = originY + (int)Math.Round(toY * dragScaleY);
-
-        MarkCursor(fx, fy);
-        SetCursorPos(fx, fy);
-        System.Threading.Thread.Sleep(60);
-        CheckCancel(p);
-        SendMouse(MOUSEEVENTF_LEFTDOWN);
-        try
-        {
-            System.Threading.Thread.Sleep(80);
-            const int steps = 14;
-            for (int i = 1; i <= steps; i++)
-            {
-                CheckCancel(p);
-                int x = fx + (tx - fx) * i / steps;
-                int y = fy + (ty - fy) * i / steps;
-                SetCursorPos(x, y);
-                System.Threading.Thread.Sleep(18);
-            }
-            System.Threading.Thread.Sleep(90);
-            CheckCancel(p);
-        }
-        finally { SendMouse(MOUSEEVENTF_LEFTUP); }
-
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["dragged"] = true;
-        r["from"] = new long[] { fx, fy };
-        r["to"] = new long[] { tx, ty };
-        return r;
-    }
-
-    public Dictionary<string, object> SetValue(Dictionary<string, object> p)
-    {
-        CheckCancel(p);
-        AssertUnlocked();
-        EnsureOverlay(8000);
-        IntPtr h = ResolveWindow(p);
-        AssertAppApproved(h);
-        long elementIndex = Json.GetLong(p, "element_index", -1);
-        if (elementIndex < 0) throw new ToolError("bad_request", "set_value requires 'element_index'");
-        string value = Json.GetString(p, "value");
-        if (value == null) throw new ToolError("bad_request", "set_value requires 'value'");
-        AutomationElement el = ElementAt(h, elementIndex);
-        object pattern;
-        if (!el.TryGetCurrentPattern(ValuePattern.Pattern, out pattern))
-            throw new ToolError("pattern_unavailable", "element [" + elementIndex + "] does not expose the Value pattern; use click + type_text instead");
-        ((ValuePattern)pattern).SetValue(value);
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["valueSet"] = true;
-        r["element_index"] = elementIndex;
-        return r;
-    }
-
-    public Dictionary<string, object> PerformSecondaryAction(Dictionary<string, object> p)
-    {
-        CheckCancel(p);
-        AssertUnlocked();
-        EnsureOverlay(8000);
-        IntPtr h = ResolveWindow(p);
-        AssertAppApproved(h);
-        long elementIndex = Json.GetLong(p, "element_index", -1);
-        string action = Json.GetString(p, "action");
-        if (elementIndex < 0 || string.IsNullOrEmpty(action))
-            throw new ToolError("bad_request", "perform_secondary_action requires 'element_index' and 'action'");
-        AutomationElement el = ElementAt(h, elementIndex);
-        string a = action.Trim().ToLowerInvariant();
-        object pattern;
-
-        if ((a == "raise" || a == "focus") && el.TryGetCurrentPattern(InvokePattern.Pattern, out pattern))
-        { ((InvokePattern)pattern).Invoke(); }
-        else if (a == "raise" || a == "focus") { try { el.SetFocus(); } catch (Exception ex) { throw new ToolError("action_failed", ex.Message); } }
-        else if ((a == "invoke" || a == "press") && el.TryGetCurrentPattern(InvokePattern.Pattern, out pattern))
-        { ((InvokePattern)pattern).Invoke(); }
-        else if ((a == "expand" || a == "collapse") && el.TryGetCurrentPattern(ExpandCollapsePattern.Pattern, out pattern))
-        {
-            ExpandCollapsePattern ec = (ExpandCollapsePattern)pattern;
-            if (a == "expand") ec.Expand(); else ec.Collapse();
-        }
-        else if ((a == "toggle") && el.TryGetCurrentPattern(TogglePattern.Pattern, out pattern))
-        { ((TogglePattern)pattern).Toggle(); }
-        else if ((a == "select" || a == "selectitem") && el.TryGetCurrentPattern(SelectionItemPattern.Pattern, out pattern))
-        { ((SelectionItemPattern)pattern).Select(); }
-        else if (a.StartsWith("scroll") && el.TryGetCurrentPattern(ScrollPattern.Pattern, out pattern))
-        {
-            ScrollPattern sp = (ScrollPattern)pattern;
-            if (a.Contains("up")) sp.ScrollVertical(ScrollAmount.SmallDecrement);
-            else if (a.Contains("down")) sp.ScrollVertical(ScrollAmount.SmallIncrement);
-            else if (a.Contains("left")) sp.ScrollHorizontal(ScrollAmount.SmallDecrement);
-            else if (a.Contains("right")) sp.ScrollHorizontal(ScrollAmount.SmallIncrement);
-            else throw new ToolError("bad_action", "unsupported scroll direction in '" + action + "'");
-        }
-        else
-        {
-            throw new ToolError("action_unavailable", "element [" + elementIndex + "] does not expose action '" + action + "'");
-        }
-
-        Dictionary<string, object> r = new Dictionary<string, object>();
-        r["performed"] = action;
-        r["element_index"] = elementIndex;
-        return r;
-    }
-}
+
+using System;
+using System.Collections.Generic;
+using System.Diagnostics;
+using System.Drawing;
+using System.Drawing.Imaging;
+using System.IO;
+using System.Runtime.InteropServices;
+using System.Text;
+using System.Windows.Automation;
+
+// ---------------------------------------------------------------------------
+// Platform driver contract. Windows is the only implementation today; the
+// interface exists so macOS (AX + CGEvent) and Linux (AT-SPI) drivers can be
+// added without touching the DSH plugin layer or the stdio protocol.
+// ---------------------------------------------------------------------------
+public interface IPlatformDriver : IDisposable
+{
+    string PlatformName { get; }
+    Dictionary<string, object> Ping(Dictionary<string, object> p);
+    Dictionary<string, object> Hello(Dictionary<string, object> p);
+    Dictionary<string, object> RestoreCursor(Dictionary<string, object> p);
+    void Maintenance();
+    Dictionary<string, object> Close(Dictionary<string, object> p);
+    Dictionary<string, object> ListWindows(Dictionary<string, object> p);
+    Dictionary<string, object> ListApps(Dictionary<string, object> p);
+    Dictionary<string, object> GetWindow(Dictionary<string, object> p);
+    Dictionary<string, object> LaunchApp(Dictionary<string, object> p);
+    Dictionary<string, object> GetWindowState(Dictionary<string, object> p);
+    Dictionary<string, object> Click(Dictionary<string, object> p);
+    Dictionary<string, object> PressKey(Dictionary<string, object> p);
+    Dictionary<string, object> TypeText(Dictionary<string, object> p);
+    Dictionary<string, object> Scroll(Dictionary<string, object> p);
+    Dictionary<string, object> Drag(Dictionary<string, object> p);
+    Dictionary<string, object> SetValue(Dictionary<string, object> p);
+    Dictionary<string, object> PerformSecondaryAction(Dictionary<string, object> p);
+    Dictionary<string, object> ActivateWindow(Dictionary<string, object> p);
+    Dictionary<string, object> OverlayShow(Dictionary<string, object> p);
+    Dictionary<string, object> OverlayHide(Dictionary<string, object> p);
+    Dictionary<string, object> Configure(Dictionary<string, object> p);
+    Dictionary<string, object> ApproveApp(Dictionary<string, object> p);
+    Dictionary<string, object> EndTurn(Dictionary<string, object> p);
+    Dictionary<string, object> InvalidateState(Dictionary<string, object> p);
+    Dictionary<string, object> StartAudioRecording(Dictionary<string, object> p);
+    Dictionary<string, object> StopAudioRecording(Dictionary<string, object> p);
+}
+
+/// Per-request time budget. Codex sends x-oai-cua-request-budget-ms with every call
+/// and the helper honours it; here the budget bounds capture waits and is reported
+/// back when a request overruns so the client and the helper agree on the deadline.
+public static class Budget
+{
+    static long _deadlineTicks;
+    static string _cancelFile;
+
+    public static void Begin(long ms, string cancelFile)
+    {
+        _deadlineTicks = ms > 0 ? DateTime.UtcNow.Ticks + ms * TimeSpan.TicksPerMillisecond : 0;
+        _cancelFile = cancelFile;
+    }
+
+    public static void Check()
+    {
+        if (_cancelFile != null && File.Exists(_cancelFile))
+        {
+            string reason = "cancelled";
+            try { reason = File.ReadAllText(_cancelFile).Trim(); } catch { }
+            if (reason != "turn_ended" && reason != "timeout") reason = "cancelled";
+            throw new ToolError(reason, "the native request was stopped; take a fresh observation before continuing");
+        }
+        if (Expired()) throw new ToolError("timeout", "the native request exceeded its time budget");
+    }
+
+    public static long RemainingMs()
+    {
+        if (_deadlineTicks == 0) return long.MaxValue / 4;
+        return (_deadlineTicks - DateTime.UtcNow.Ticks) / TimeSpan.TicksPerMillisecond;
+    }
+
+    public static bool Expired()
+    {
+        return _deadlineTicks != 0 && RemainingMs() <= 0;
+    }
+}
+
+public class ToolError : Exception
+{
+    public readonly string Code;
+    public ToolError(string code, string message) : base(message) { Code = code; }
+}
+
+// ---------------------------------------------------------------------------
+// Windows driver: Win32 + UI Automation + PrintWindow.
+// ---------------------------------------------------------------------------
+public class WindowsDriver : IPlatformDriver
+{
+    public string PlatformName { get { return "windows"; } }
+
+    // ---- win32 ----
+    delegate bool EnumWindowsProc(IntPtr hWnd, IntPtr lParam);
+
+    [DllImport("user32.dll")] static extern bool EnumWindows(EnumWindowsProc cb, IntPtr lParam);
+    [DllImport("user32.dll")] static extern bool IsWindowVisible(IntPtr hWnd);
+    [DllImport("user32.dll")] static extern bool IsWindow(IntPtr hWnd);
+    [DllImport("user32.dll", CharSet = CharSet.Unicode)] static extern int GetWindowTextW(IntPtr hWnd, StringBuilder sb, int max);
+    [DllImport("user32.dll", CharSet = CharSet.Unicode)] static extern int GetClassNameW(IntPtr hWnd, StringBuilder sb, int max);
+    [DllImport("user32.dll")] static extern uint GetWindowThreadProcessId(IntPtr hWnd, out uint pid);
+    [DllImport("user32.dll")] static extern bool GetWindowRect(IntPtr hWnd, out RECT r);
+    [DllImport("user32.dll")] static extern bool GetClientRect(IntPtr hWnd, out RECT r);
+    [DllImport("user32.dll")] static extern bool PrintWindow(IntPtr hWnd, IntPtr hdc, uint flags);
+    [DllImport("user32.dll")] static extern bool SetForegroundWindow(IntPtr hWnd);
+    [DllImport("user32.dll")] static extern bool ShowWindow(IntPtr hWnd, int cmd);
+    [DllImport("user32.dll")] static extern bool IsIconic(IntPtr hWnd);
+    [DllImport("user32.dll")] static extern bool SetCursorPos(int x, int y);
+    [DllImport("user32.dll")] static extern bool GetCursorPos(out POINT p);
+    [DllImport("user32.dll")] static extern uint SendInput(uint n, INPUT[] inputs, int size);
+    [DllImport("user32.dll")] static extern IntPtr GetForegroundWindow();
+    [DllImport("user32.dll")] static extern IntPtr SetFocus(IntPtr hWnd);
+    [DllImport("user32.dll")] static extern bool AttachThreadInput(uint idAttach, uint idAttachTo, bool fAttach);
+    [DllImport("kernel32.dll")] static extern uint GetCurrentThreadId();
+    [DllImport("user32.dll")] static extern uint GetWindowThreadProcessId(IntPtr hWnd, IntPtr pid);
+    [DllImport("user32.dll")] static extern bool IsWindowEnabled(IntPtr hWnd);
+    [DllImport("user32.dll")] static extern bool SystemParametersInfo(uint action, uint param, IntPtr pv, uint winIni);
+    [DllImport("user32.dll")] static extern bool SwitchToThisWindow(IntPtr hWnd, bool fAltTab);
+    [DllImport("user32.dll")] static extern void keybd_event(byte bVk, byte bScan, uint dwFlags, UIntPtr dwExtraInfo);
+    [DllImport("user32.dll")] static extern bool BringWindowToTop(IntPtr hWnd);
+    [DllImport("user32.dll")] static extern IntPtr SetThreadDpiAwarenessContext(IntPtr ctx);
+    [DllImport("user32.dll")] static extern int GetSystemMetrics(int index);
+    [DllImport("user32.dll")] static extern bool SetProcessDPIAware();
+    [DllImport("dwmapi.dll")] static extern int DwmGetWindowAttribute(IntPtr hwnd, int attr, out RECT r, int size);
+    [DllImport("user32.dll")] static extern IntPtr GetWindow(IntPtr hWnd, uint cmd);
+    [DllImport("user32.dll")] static extern int GetWindowLong(IntPtr hWnd, int index);
+
+    [StructLayout(LayoutKind.Sequential)] struct RECT { public int Left, Top, Right, Bottom; }
+    [StructLayout(LayoutKind.Sequential)] struct POINT { public int X, Y; }
+
+    [StructLayout(LayoutKind.Sequential)]
+    struct MOUSEINPUT { public int dx; public int dy; public uint mouseData; public uint dwFlags; public uint time; public IntPtr dwExtraInfo; }
+
+    [StructLayout(LayoutKind.Sequential)]
+    struct KEYBDINPUT { public ushort wVk; public ushort wScan; public uint dwFlags; public uint time; public IntPtr dwExtraInfo; }
+
+    [StructLayout(LayoutKind.Explicit)]
+    struct INPUTUNION
+    {
+        [FieldOffset(0)] public MOUSEINPUT mi;
+        [FieldOffset(0)] public KEYBDINPUT ki;
+    }
+
+    [StructLayout(LayoutKind.Sequential)]
+    struct INPUT { public uint type; public INPUTUNION u; }
+
+    const int INPUT_MOUSE = 0;
+    const int INPUT_KEYBOARD = 1;
+    const uint KEYEVENTF_KEYUP = 0x0002;
+    const uint KEYEVENTF_UNICODE = 0x0004;
+    const uint KEYEVENTF_EXTENDEDKEY = 0x0001;
+    const uint MOUSEEVENTF_LEFTDOWN = 0x0002;
+    const uint MOUSEEVENTF_LEFTUP = 0x0004;
+    const uint MOUSEEVENTF_RIGHTDOWN = 0x0008;
+    const uint MOUSEEVENTF_RIGHTUP = 0x0010;
+    const uint MOUSEEVENTF_MIDDLEDOWN = 0x0020;
+    const uint MOUSEEVENTF_MIDDLEUP = 0x0040;
+    const uint MOUSEEVENTF_WHEEL = 0x0800;
+    const uint MOUSEEVENTF_HWHEEL = 0x1000;
+    const int DWMWA_EXTENDED_FRAME_BOUNDS = 9;
+
+    // ---- state ----
+    readonly object _sync = new object();
+    sealed class Snapshot
+    {
+        public long WindowId;
+        public DateTime At;
+        public Dictionary<long, AutomationElement> Elements = new Dictionary<long, AutomationElement>();
+        public Dictionary<long, string> FallbackLabels = new Dictionary<long, string>();
+        public List<object> Summary = new List<object>();
+    }
+
+    readonly Dictionary<long, Snapshot> _snapshots = new Dictionary<long, Snapshot>();
+    int _snapshotTtlMs = 120000;
+
+    /// A screenshot region the model was handed, so a later coordinate action can
+    /// be interpreted in that exact screenshot space (Codex: screenshotId).
+    sealed class ShotRegion
+    {
+        public long WindowId;
+        public int OriginX;
+        public int OriginY;
+        public int Width;
+        public int Height;
+        /// Physical pixels per screenshot pixel. 1.0 for DPI-aware targets; >1 when
+        /// the target is DPI-virtualized and PrintWindow returns an unscaled render.
+        public double ScaleX = 1.0;
+        public double ScaleY = 1.0;
+        public DateTime At;
+    }
+
+    // In-use indicator ("DSH 正在使用您的电脑" + whale-girl). Codex draws an animated
+    // border and a virtual cursor; this is the DSH equivalent, and Esc is the kill
+    // switch: the overlay drops a cancel file that every input path checks first.
+    Process _overlay;
+    System.Threading.Timer _cursorWatchdog;
+    DateTime _overlayUntil = DateTime.MinValue;
+    string _overlayCancelFile;
+    string _overlayText = "DSH 正在使用您的电脑";
+    string _overlayHint = "Esc 取消";
+    string _overlayAccent = "#3f7fd8";
+    string _cursorFile;      // the overlay polls this to draw the virtual cursor
+    /// Turn token that the user interrupted with Esc. Codex persists the same
+    /// fact as a marker file per conversation+turn; here the token travels with
+    /// every input call, so the block ends exactly when the turn does.
+    string _interruptedTurn = null;
+    /// Apps the plugin approved through the DSH approval seam. Codex passes the same
+    /// fact per request as x-oai-cua-approved-app; here the helper is the enforcer.
+    readonly HashSet<string> _approvedApps = new HashSet<string>(StringComparer.OrdinalIgnoreCase);
+    bool _requireAppApproval = false;
+    Process _audioProcess;
+    string _audioPath;
+    string _audioStopFile;
+
+    readonly Dictionary<string, ShotRegion> _shots = new Dictionary<string, ShotRegion>();
+    int _shotTtlMs = 120000;
+    int _shotSeq = 0;
+    readonly string _shotGeneration = Guid.NewGuid().ToString("N");
+
+    const uint GW_OWNER = 4;
+    const int GWL_STYLE = -16;
+    const int WS_POPUP = unchecked((int)0x80000000);
+
+    public WindowsDriver()
+    {
+        Maintenance(); // startup repair: a previous session may have died with cursors hidden
+        try { SetThreadDpiAwarenessContext(new IntPtr(-4)); } catch { }
+        try { SetProcessDPIAware(); } catch { }
+    }
+
+    public void Dispose()
+    {
+        if (_cursorWatchdog != null) { _cursorWatchdog.Dispose(); _cursorWatchdog = null; }
+        StopOverlay();
+    }
+
+    // ---- approval plumbing ----
+    public Dictionary<string, object> Configure(Dictionary<string, object> p)
+    {
+        _requireAppApproval = Json.GetBool(p, "requireAppApproval", _requireAppApproval);
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["requireAppApproval"] = _requireAppApproval;
+        return r;
+    }
+
+    public Dictionary<string, object> ApproveApp(Dictionary<string, object> p)
+    {
+        string app = Json.GetString(p, "app");
+        if (string.IsNullOrEmpty(app)) throw new ToolError("bad_request", "approve_app requires 'app'");
+        lock (_sync) { _approvedApps.Add(app); }
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["approved"] = app;
+        r["count"] = (long)_approvedApps.Count;
+        return r;
+    }
+
+    /// Turn boundary: drop per-turn state (element snapshots, screenshot regions, the
+    /// Esc block, the indicator) but keep session-scoped app approvals.
+    public Dictionary<string, object> EndTurn(Dictionary<string, object> p)
+    {
+        string ending = Json.GetString(p, "turnToken");
+        if (ending != null) ClearInterrupt(ending);
+        lock (_sync) { _snapshots.Clear(); _shots.Clear(); }
+        _interruptedTurn = null;
+        StopOverlay();
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["ended"] = true;
+        r["approvedApps"] = (long)_approvedApps.Count;
+        return r;
+    }
+
+    public Dictionary<string, object> InvalidateState(Dictionary<string, object> p)
+    {
+        lock (_sync) { _snapshots.Clear(); _shots.Clear(); }
+        StopOverlay();
+        return new Dictionary<string, object> { { "invalidated", true } };
+    }
+
+    void AssertAppApproved(IntPtr h)
+    {
+        if (!_requireAppApproval) return;
+        uint pid;
+        GetWindowThreadProcessId(h, out pid);
+        string id = ProcessPathOf(pid);
+        lock (_sync) { if (_approvedApps.Contains(id)) return; }
+        throw new ToolError("app_not_approved", "the Computer Use policy has not approved '" + id + "' yet. Approve the app through the DSH approval prompt before acting on its windows.");
+    }
+
+    // ---- audio recording (WASAPI loopback) ----
+    public Dictionary<string, object> StartAudioRecording(Dictionary<string, object> p)
+    {
+        bool envEnabled = Environment.GetEnvironmentVariable("DSH_CU_AUDIO") == "1";
+        bool force = Json.GetBool(p, "force", false);
+        // Codex treats computer audio as its own app ("computer-audio") and asks for a
+        // separate approval before recording; the plugin passes that here.
+        bool audioApproved = Json.GetString(p, "approvedApp") == "computer-audio";
+        if (!envEnabled && !force && !audioApproved)
+            throw new ToolError("audio_not_enabled", "computer audio recording is not enabled. Approve it (app \"computer-audio\") or set DSH_CU_AUDIO=1.");
+        if (_audioProcess != null && !_audioProcess.HasExited)
+            throw new ToolError("audio_running", "an audio recording is already running; stop it first");
+
+        string dir = Path.GetDirectoryName(typeof(WindowsDriver).Assembly.Location) ?? ".";
+        string exe = Path.Combine(dir, "audio", "dsh-cu-audio.exe");
+        if (!File.Exists(exe)) throw new ToolError("audio_missing", "audio recorder not found at " + exe);
+
+        long maxMs = Json.GetLong(p, "maxDurationMs", 60000);
+        if (maxMs < 500) maxMs = 500;
+        if (maxMs > 300000) maxMs = 300000;   // Codex caps a single recording at 5 minutes
+        _audioPath = Path.Combine(Path.GetTempPath(), "dsh-cu-audio-" + Guid.NewGuid().ToString("N") + ".wav");
+        _audioStopFile = Path.Combine(Path.GetTempPath(), "dsh-cu-audio-stop-" + Guid.NewGuid().ToString("N") + ".flag");
+        try { if (File.Exists(_audioStopFile)) File.Delete(_audioStopFile); } catch { }
+
+        ProcessStartInfo psi = new ProcessStartInfo(exe, "--out \"" + _audioPath + "\" --max-ms " + maxMs + " --stop-file \"" + _audioStopFile + "\"");
+        psi.UseShellExecute = false;
+        psi.CreateNoWindow = true;
+        psi.RedirectStandardOutput = true;
+        psi.RedirectStandardError = true;
+        _audioProcess = Process.Start(psi);
+        System.Threading.Thread.Sleep(400);
+        if (_audioProcess.HasExited)
+            throw new ToolError("audio_not_started", "computer audio recording did not start in time (the recorder exited immediately)");
+
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["started"] = true;
+        r["filepath"] = _audioPath;
+        r["maxDurationMs"] = maxMs;
+        return r;
+    }
+
+    public Dictionary<string, object> StopAudioRecording(Dictionary<string, object> p)
+    {
+        if (_audioProcess == null) throw new ToolError("audio_not_started", "no audio recording is running");
+        try { if (_audioStopFile != null) File.WriteAllText(_audioStopFile, "stop"); } catch { }
+        try { _audioProcess.WaitForExit(8000); } catch { }
+        try { if (_audioProcess != null && !_audioProcess.HasExited) _audioProcess.Kill(); } catch { }
+
+        string path = _audioPath;
+        _audioProcess = null;
+        _audioStopFile = null;
+        _audioPath = null;
+
+        byte[] audio = new byte[0];
+        try { if (path != null && File.Exists(path)) audio = File.ReadAllBytes(path); } catch { }
+
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["filepath"] = path;
+        // Codex returns the raw WAV bytes plus a data URL; base64 is the JSON-safe form.
+        r["bytes"] = Convert.ToBase64String(audio);
+        r["byteLength"] = (long)audio.Length;
+        r["data_url"] = audio.Length > 44 ? "data:audio/wav;base64," + Convert.ToBase64String(audio) : "";
+        r["ok"] = audio.Length > 44;
+        if (audio.Length <= 44) throw new ToolError("audio_not_recorded", "computer audio recording produced no samples");
+        return r;
+    }
+
+    // ---- in-use indicator ----
+    static string OverlayExePath()
+    {
+        string dir = Path.GetDirectoryName(typeof(WindowsDriver).Assembly.Location);
+        return Path.Combine(dir ?? ".", "dsh-cu-overlay.exe");
+    }
+
+    public Dictionary<string, object> OverlayShow(Dictionary<string, object> p)
+    {
+        string text = Json.GetString(p, "text");
+        if (!string.IsNullOrEmpty(text)) _overlayText = text;
+        string hint = Json.GetString(p, "hint");
+        if (!string.IsNullOrEmpty(hint)) _overlayHint = hint;
+        string accent = Json.GetString(p, "accent");
+        if (!string.IsNullOrEmpty(accent)) _overlayAccent = accent;
+        EnsureOverlay(Json.GetLong(p, "ttlMs", 20000));
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["visible"] = true;
+        r["cancelFile"] = _overlayCancelFile;
+        return r;
+    }
+
+    public Dictionary<string, object> OverlayHide(Dictionary<string, object> p)
+    {
+        StopOverlay();
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["visible"] = false;
+        return r;
+    }
+
+    void EnsureOverlay(long ttlMs)
+    {
+        string exe = OverlayExePath();
+        if (!File.Exists(exe)) return;
+        string mode = Environment.GetEnvironmentVariable("DSH_CU_INDICATOR");
+        if (mode != null && mode.Equals("off", StringComparison.OrdinalIgnoreCase)) { StopOverlay(); return; }
+        string ttlEnv = Environment.GetEnvironmentVariable("DSH_CU_INDICATOR_TTL_MS");
+        long parsed;
+        if (ttlEnv != null && long.TryParse(ttlEnv, out parsed) && parsed >= 1000) ttlMs = parsed;
+        if (_overlay != null && !_overlay.HasExited && DateTime.UtcNow < _overlayUntil)
+        {
+            _overlayUntil = DateTime.UtcNow.AddMilliseconds(ttlMs);
+            return;
+        }
+        StopOverlay();
+        _overlayCancelFile = Path.Combine(Path.GetTempPath(), "dsh-cu-cancel-" + Guid.NewGuid().ToString("N") + ".flag");
+        if (_cursorFile == null) _cursorFile = Path.Combine(Path.GetTempPath(), "dsh-cu-cursor-" + Guid.NewGuid().ToString("N").Substring(0, 8) + ".flag");
+        try { if (File.Exists(_overlayCancelFile)) File.Delete(_overlayCancelFile); } catch { }
+        StringBuilder args = new StringBuilder();
+        args.Append("--ttl-ms ").Append(ttlMs);
+        args.Append(" --cancel-file \"").Append(_overlayCancelFile).Append('"');
+        args.Append(" --text \"").Append(_overlayText).Append('"');
+        args.Append(" --hint \"").Append(_overlayHint).Append('"');
+        args.Append(" --owner-pid ").Append(Process.GetCurrentProcess().Id);
+        args.Append(" --accent \"").Append(_overlayAccent).Append('"');
+        if (_cursorFile != null) args.Append(" --cursor-file \"").Append(_cursorFile).Append('"');
+        // Suppression is the default (the virtual cursor stands in for the real one);
+        // DSH_CU_KEEP_CURSOR=1 keeps the real pointer visible instead.
+        if (Environment.GetEnvironmentVariable("DSH_CU_KEEP_CURSOR") == "1") args.Append(" --no-suppress-cursor");
+        try
+        {
+            ProcessStartInfo psi = new ProcessStartInfo(exe, args.ToString());
+            psi.UseShellExecute = false;
+            _overlay = Process.Start(psi);
+            _overlayUntil = DateTime.UtcNow.AddMilliseconds(ttlMs);
+            // Codex runs a dedicated system-cursor-manager process that restores the
+            // pointer the moment its parent dies. This timer gives the same guarantee
+            // without a second exe: if the overlay vanishes with cursors hidden, the
+            // pointer comes back within half a second, with no request needed.
+            if (_cursorWatchdog == null)
+                _cursorWatchdog = new System.Threading.Timer(delegate { try { Maintenance(); } catch { } }, null, 500, 500);
+        }
+        catch { _overlay = null; }
+    }
+
+    void StopOverlay()
+    {
+        try { if (_overlay != null && !_overlay.HasExited) _overlay.Kill(); } catch { }
+        // Kill() skips the overlay's exit handlers, which is where it would restore the
+        // system cursors. Restore them from here so a terminated overlay can never leave
+        // the user without a pointer.
+        RestoreCursors();
+        try { File.Delete(OverlayCursorFlagPath()); } catch { }
+        try { if (_overlayCancelFile != null && File.Exists(_overlayCancelFile)) File.Delete(_overlayCancelFile); } catch { }
+        _overlay = null;
+        _overlayUntil = DateTime.MinValue;
+    }
+
+    /// Esc on the overlay drops a cancel file; honour it before any other input.
+    const string EscMessage = "Computer Use was stopped by the user with the physical Escape key. Stop your work, do not call further Computer Use tools in this turn, and send a final message noting that the user stopped Computer Use.";
+
+    void CheckCancel(Dictionary<string, object> p)
+    {
+        Budget.Check();
+        string token = p == null ? null : Json.GetString(p, "turnToken");
+        if (token != null && (token == _interruptedTurn || InterruptMarkerExists(token)))
+            throw new ToolError("turn_stopped", EscMessage);
+        bool overlayExited = _overlay != null && _overlay.HasExited && _overlay.ExitCode == 130;
+        bool cancelFile = _overlayCancelFile != null && File.Exists(_overlayCancelFile);
+        if (!overlayExited && !cancelFile) return;
+        if (cancelFile) { try { File.Delete(_overlayCancelFile); } catch { } }
+        _interruptedTurn = token;
+        PersistInterrupt(token);
+        StopOverlay();
+        Program.RequestedExitCode = 130;   // Codex exits 130 when the user hits Esc
+        throw new ToolError("cancelled_by_user", EscMessage);
+    }
+
+    const int ProtocolVersion = 2;
+
+    static string InterruptMarkerPath(string token)
+    {
+        string root = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "dsh-computer-use", "interrupts");
+        StringBuilder safe = new StringBuilder();
+        if (token != null)
+        {
+            for (int i = 0; i < token.Length && safe.Length < 64; i++)
+            {
+                char ch = token[i];
+                if (char.IsLetterOrDigit(ch) || ch == '-' || ch == '_' || ch == '.') safe.Append(ch);
+                else if (ch == ':') safe.Append('-');
+            }
+        }
+        if (safe.Length == 0) safe.Append("turn");
+        return Path.Combine(root, safe.ToString() + ".flag");
+    }
+
+    static void PersistInterrupt(string token)
+    {
+        try
+        {
+            string path = InterruptMarkerPath(token);
+            Directory.CreateDirectory(Path.GetDirectoryName(path));
+            File.WriteAllText(path, DateTime.UtcNow.ToString("o"));
+        }
+        catch { }
+    }
+
+    static bool InterruptMarkerExists(string token)
+    {
+        try { return File.Exists(InterruptMarkerPath(token)); } catch { return false; }
+    }
+
+    static void ClearInterrupt(string token)
+    {
+        try { File.Delete(InterruptMarkerPath(token)); } catch { }
+    }
+
+    /// Publish the pointer target for the virtual cursor. The overlay reads this
+    /// file; a missing file simply means "no cursor to draw".
+    void MarkCursor(int x, int y)
+    {
+        if (_cursorFile == null) return;
+        try { File.WriteAllText(_cursorFile, x.ToString(System.Globalization.CultureInfo.InvariantCulture) + "," + y.ToString(System.Globalization.CultureInfo.InvariantCulture) + "," + DateTime.UtcNow.Ticks.ToString(System.Globalization.CultureInfo.InvariantCulture)); } catch { }
+    }
+
+    /// Codex reports screenLocked instead of failing with an opaque automation error.
+    static void AssertUnlocked()
+    {
+        IntPtr fg = GetForegroundWindow();
+        if (fg == IntPtr.Zero) return;
+        uint pid;
+        GetWindowThreadProcessId(fg, out pid);
+        string proc = "";
+        try { proc = Process.GetProcessById((int)pid).ProcessName; } catch { }
+        if (proc.Equals("LockApp", StringComparison.OrdinalIgnoreCase) || proc.Equals("LogonUI", StringComparison.OrdinalIgnoreCase))
+            throw new ToolError("screen_locked", "the Windows desktop is locked (LockApp is in the foreground), so Computer Use cannot interact with it. Ask the user to unlock the desktop and stop trying until they do.");
+    }
+
+    // ---- helpers ----
+    static string TitleOf(IntPtr h)
+    {
+        StringBuilder sb = new StringBuilder(1024);
+        GetWindowTextW(h, sb, sb.Capacity);
+        return sb.ToString();
+    }
+
+    static string ClassOf(IntPtr h)
+    {
+        StringBuilder sb = new StringBuilder(256);
+        GetClassNameW(h, sb, sb.Capacity);
+        return sb.ToString();
+    }
+
+    static RECT BoundsOf(IntPtr h)
+    {
+        RECT r;
+        try
+        {
+            if (DwmGetWindowAttribute(h, DWMWA_EXTENDED_FRAME_BOUNDS, out r, Marshal.SizeOf(typeof(RECT))) == 0 && r.Right > r.Left)
+                return r;
+        }
+        catch { }
+        GetWindowRect(h, out r);
+        return r;
+    }
+
+    static Dictionary<string, object> RectDict(int x, int y, int w, int h)
+    {
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["x"] = x; d["y"] = y; d["width"] = w; d["height"] = h;
+        return d;
+    }
+
+    static string ProcessPathOf(uint pid)
+    {
+        try
+        {
+            Process p = Process.GetProcessById((int)pid);
+            try { return p.MainModule.FileName; }
+            catch { return p.ProcessName; }
+        }
+        catch { return "unknown"; }
+    }
+
+    static string ProcessNameOf(uint pid)
+    {
+        try { return Process.GetProcessById((int)pid).ProcessName; }
+        catch { return "unknown"; }
+    }
+
+    Dictionary<string, object> AppOf(uint pid)
+    {
+        Dictionary<string, object> app = new Dictionary<string, object>();
+        string full = ProcessPathOf(pid);
+        app["id"] = full;
+        app["name"] = Path.GetFileNameWithoutExtension(full);
+        app["displayName"] = ProcessNameOf(pid);
+        app["pid"] = (long)pid;
+        try
+        {
+            Process proc = Process.GetProcessById((int)pid);
+            app["lastUsedDate"] = proc.StartTime.ToUniversalTime().ToString("o");
+        }
+        catch { }
+        return app;
+    }
+
+    Dictionary<string, object> WindowDict(IntPtr h, uint pid, int z)
+    {
+        RECT r = BoundsOf(h);
+        Dictionary<string, object> w = new Dictionary<string, object>();
+        w["id"] = (long)h.ToInt64();
+        w["title"] = TitleOf(h);
+        w["app"] = AppOf(pid);
+        w["bounds"] = RectDict(r.Left, r.Top, r.Right - r.Left, r.Bottom - r.Top);
+        w["zIndex"] = z;
+        w["className"] = ClassOf(h);
+        return w;
+    }
+
+    List<IntPtr> EnumerateWindows(bool visibleOnly)
+    {
+        List<IntPtr> list = new List<IntPtr>();
+        EnumWindows(delegate(IntPtr h, IntPtr lp)
+        {
+            if (!IsWindow(h)) return true;
+            if (visibleOnly && !IsWindowVisible(h)) return true;
+            uint pid;
+            GetWindowThreadProcessId(h, out pid);
+            if (pid == 0) return true;
+            RECT r;
+            GetWindowRect(h, out r);
+            if (r.Right - r.Left < 80 || r.Bottom - r.Top < 60) return true;
+            string title = TitleOf(h);
+            if (title.Length == 0) return true;
+            list.Add(h);
+            return true;
+        }, IntPtr.Zero);
+        return list;
+    }
+
+    // ---- methods ----
+    public Dictionary<string, object> Ping(Dictionary<string, object> p)
+    {
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["platform"] = PlatformName;
+        d["pid"] = (long)Process.GetCurrentProcess().Id;
+        d["overlayPid"] = _overlay != null && !_overlay.HasExited ? (object)(long)_overlay.Id : null;
+        d["ok"] = true;
+        d["uia"] = true;
+        return d;
+    }
+
+    /// Run cheaply before every request: if a previous overlay died while it was hiding
+    /// the system cursors, put them back. A blank desktop cursor is worse than a stray
+    /// visible one, so this never assumes the overlay cleaned up after itself.
+    public void Maintenance()
+    {
+        try
+        {
+            string flag = OverlayCursorFlagPath();
+            if (!File.Exists(flag)) return;
+            int pid;
+            if (!int.TryParse(File.ReadAllText(flag).Trim(), out pid)) { RestoreCursors(); return; }
+            bool alive;
+            try { Process.GetProcessById(pid); alive = true; } catch { alive = false; }
+            if (!alive) { RestoreCursors(); try { File.Delete(flag); } catch { } }
+        }
+        catch { }
+    }
+
+    static string OverlayCursorFlagPath()
+    {
+        return Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.LocalApplicationData), "dsh-computer-use", "cursor-hidden.flag");
+    }
+
+    static void RestoreCursors()
+    {
+        try { SystemParametersInfo(0x0057, 0, IntPtr.Zero, 0); } catch { }
+    }
+
+    public Dictionary<string, object> RestoreCursor(Dictionary<string, object> p)
+    {
+        RestoreCursors();
+        try { File.Delete(OverlayCursorFlagPath()); } catch { }
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["restored"] = true;
+        return r;
+    }
+
+    public Dictionary<string, object> Hello(Dictionary<string, object> p)
+    {
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["ok"] = true;
+        d["name"] = "dsh-computer-use";
+        d["protocolVersion"] = ProtocolVersion;
+        d["platform"] = PlatformName;
+        List<object> caps = new List<object>();
+        foreach (string c in new string[] {
+            "window2", "accessibility_tree", "multi_screenshot", "wgc_capture",
+            "virtual_cursor", "audio_recording", "turn_interrupt", "app_approval",
+            "screenshot_ids", "budgeted_capture" }) caps.Add(c);
+        d["capabilities"] = caps;
+        return d;
+    }
+
+    public Dictionary<string, object> Close(Dictionary<string, object> p)
+    {
+        Program.RequestedExitCode = 0;
+        Program.CloseRequested = true;
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["closed"] = true;
+        return d;
+    }
+
+    public Dictionary<string, object> ListWindows(Dictionary<string, object> p)
+    {
+        List<object> outList = new List<object>();
+        List<IntPtr> hs = EnumerateWindows(true);
+        for (int i = 0; i < hs.Count; i++)
+        {
+            uint pid;
+            GetWindowThreadProcessId(hs[i], out pid);
+            outList.Add(WindowDict(hs[i], pid, i));
+        }
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["windows"] = outList;
+        d["count"] = outList.Count;
+        return d;
+    }
+
+    public Dictionary<string, object> ListApps(Dictionary<string, object> p)
+    {
+        Dictionary<string, List<object>> grouped = new Dictionary<string, List<object>>();
+        Dictionary<string, Dictionary<string, object>> appInfo = new Dictionary<string, Dictionary<string, object>>();
+        List<IntPtr> hs = EnumerateWindows(true);
+        for (int i = 0; i < hs.Count; i++)
+        {
+            uint pid;
+            GetWindowThreadProcessId(hs[i], out pid);
+            Dictionary<string, object> app = AppOf(pid);
+            string key = Convert.ToString(app["id"]);
+            if (!grouped.ContainsKey(key)) { grouped[key] = new List<object>(); appInfo[key] = app; }
+            grouped[key].Add(WindowDict(hs[i], pid, i));
+        }
+        List<object> apps = new List<object>();
+        foreach (KeyValuePair<string, List<object>> kv in grouped)
+        {
+            Dictionary<string, object> a = appInfo[kv.Key];
+            a["isRunning"] = true;
+            try
+            {
+                IntPtr sample = IntPtr.Zero;
+                List<IntPtr> windowsOfApp = new List<IntPtr>();
+                for (int w = 0; w < hs.Count; w++)
+                {
+                    uint wpid;
+                    GetWindowThreadProcessId(hs[w], out wpid);
+                    if (Convert.ToString(AppOf(wpid)["id"]) == kv.Key) { sample = hs[w]; break; }
+                }
+                if (sample != IntPtr.Zero)
+                {
+                    uint spid;
+                    GetWindowThreadProcessId(sample, out spid);
+                    a["lastUsedDate"] = Process.GetProcessById((int)spid).StartTime.ToUniversalTime().ToString("o");
+                }
+            }
+            catch { }
+            a["windows"] = kv.Value;
+            apps.Add(a);
+        }
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["apps"] = apps;
+        d["count"] = apps.Count;
+        return d;
+    }
+
+    public Dictionary<string, object> GetWindow(Dictionary<string, object> p)
+    {
+        IntPtr h = ResolveWindow(p);
+        uint pid;
+        GetWindowThreadProcessId(h, out pid);
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["window"] = WindowDict(h, pid, 0);
+        return d;
+    }
+
+    public Dictionary<string, object> ActivateWindow(Dictionary<string, object> p)
+    {
+        IntPtr h = ResolveWindow(p);
+        Activate(h);
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["activated"] = true;
+        d["window"] = WindowDict(h, PidOf(h), 0);
+        return d;
+    }
+
+    static uint PidOf(IntPtr h) { uint pid; GetWindowThreadProcessId(h, out pid); return pid; }
+
+    /// Bring a window to the foreground and PROVE it got there. Input is never
+    /// injected into a window that is not foreground: SendInput targets whatever
+    /// holds focus, so an unproven activation would type into the user's window.
+    static void Activate(IntPtr h)
+    {
+        if (!EnsureForeground(h))
+            throw new ToolError("focus_failed", "could not bring window " + h.ToInt64() + " to the foreground, so the input was refused instead of landing in another window. " + DescribeForeground());
+    }
+
+    static bool EnsureForeground(IntPtr h)
+    {
+        if (!IsWindow(h)) throw new ToolError("window_gone", "window " + h.ToInt64() + " no longer exists; call list_windows again");
+        if (!IsWindowEnabled(h)) throw new ToolError("window_disabled", "window " + h.ToInt64() + " is disabled (a modal dialog may be blocking it)");
+        if (IsIconic(h)) { ShowWindow(h, 9); System.Threading.Thread.Sleep(180); }
+        if (GetForegroundWindow() == h) return true;
+
+        uint fgThread = GetWindowThreadProcessId(GetForegroundWindow(), IntPtr.Zero);
+        uint thisThread = GetCurrentThreadId();
+        bool attached = false;
+        try
+        {
+            if (fgThread != 0 && fgThread != thisThread) attached = AttachThreadInput(thisThread, fgThread, true);
+            ShowWindow(h, 5); // SW_SHOW
+            BringWindowToTop(h);
+            SetForegroundWindow(h);
+            SetFocus(h);
+        }
+        finally { if (attached) AttachThreadInput(thisThread, fgThread, false); }
+
+        if (WaitForeground(h, 10)) return true;
+
+        // Windows refuses foreground changes from a background process. The
+        // documented-looking escape hatches, in order of intrusiveness:
+        //   1. SwitchToThisWindow (alt-tab semantics, no key injection)
+        //   2. a bare Alt press, which lifts the foreground lock for this call
+        try { SwitchToThisWindow(h, true); } catch { }
+        if (WaitForeground(h, 10)) return true;
+
+        keybd_event(0x12, 0, 0, UIntPtr.Zero);        // VK_MENU down
+        keybd_event(0x12, 0, 0x0002, UIntPtr.Zero);   // VK_MENU up
+        SetForegroundWindow(h);
+        if (WaitForeground(h, 12)) return true;
+        return false;
+    }
+
+    /// Name the process that currently owns focus: the usual reason a
+    /// background host cannot steal it is an always-on-top utility or an
+    /// elevated window.
+    static string DescribeForeground()
+    {
+        IntPtr fg = GetForegroundWindow();
+        if (fg == IntPtr.Zero) return "No window currently holds the foreground.";
+        uint pid;
+        GetWindowThreadProcessId(fg, out pid);
+        string proc = "unknown";
+        try { proc = Process.GetProcessById((int)pid).ProcessName; } catch { }
+        string title = TitleOf(fg);
+        string cls = ClassOf(fg);
+        return "Foreground is held by " + proc + " (pid " + pid + ", class " + cls + (title.Length > 0 ? ", title \"" + Trim(title, 60) + "\"" : "") + "). Bring the target window forward yourself, or close/pause the utility that keeps stealing focus.";
+    }
+
+    static bool WaitForeground(IntPtr h, int attempts)
+    {
+        for (int i = 0; i < attempts; i++)
+        {
+            if (GetForegroundWindow() == h) { System.Threading.Thread.Sleep(70); return true; }
+            System.Threading.Thread.Sleep(40);
+        }
+        return GetForegroundWindow() == h;
+    }
+
+    static void Slow() { System.Threading.Thread.Sleep(120); }
+
+    public Dictionary<string, object> LaunchApp(Dictionary<string, object> p)
+    {
+        string app = Json.GetString(p, "app");
+        if (string.IsNullOrEmpty(app)) throw new ToolError("bad_request", "launch_app requires 'app'");
+        // Resolution order matches the API contract: an explicit .exe path first,
+        // then a packaged-app id (AUMID) through shell:AppsFolder, then a shell URI
+        // such as ms-settings:. A bare unknown string is tried verbatim last.
+        string resolved = app;
+        if (!File.Exists(app) && app.IndexOf('\\') < 0 && !app.EndsWith(".exe", StringComparison.OrdinalIgnoreCase) && app.IndexOf(':') < 0)
+            resolved = "shell:AppsFolder\\" + app;
+        ProcessStartInfo psi = new ProcessStartInfo();
+        psi.FileName = resolved;
+        psi.UseShellExecute = true;
+        List<object> args = Json.Get(p, "args") as List<object>;
+        if (args != null && args.Count > 0)
+        {
+            StringBuilder argLine = new StringBuilder();
+            for (int i = 0; i < args.Count; i++)
+            {
+                if (i > 0) argLine.Append(' ');
+                string a = Convert.ToString(args[i], System.Globalization.CultureInfo.InvariantCulture) ?? "";
+                // A quote or newline would let a caller smuggle extra arguments past the
+                // single quoting rule below, so refuse instead of guessing.
+                if (a.IndexOf('"') >= 0 || a.IndexOf('\n') >= 0 || a.IndexOf('\r') >= 0)
+                    throw new ToolError("bad_request", "launch_app argument " + i + " contains a quote or newline; pass a single plain argument");
+                if (a.Length == 0 || a.IndexOf(' ') >= 0 || a.IndexOf('\t') >= 0)
+                    argLine.Append('"').Append(a).Append('"');
+                else argLine.Append(a);
+            }
+            psi.Arguments = argLine.ToString();
+        }
+        Process proc = Process.Start(psi);
+        System.Threading.Thread.Sleep(900);
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["launched"] = true;
+        d["app"] = app;
+        try { d["pid"] = (long)proc.Id; } catch { }
+        return d;
+    }
+
+    IntPtr ResolveWindow(Dictionary<string, object> p)
+    {
+        Dictionary<string, object> w = Json.AsObject(Json.Get(p, "window"));
+        long id = Json.GetLong(w, "id", 0);
+        if (id == 0) id = Json.GetLong(p, "windowId", 0);
+        if (id == 0) throw new ToolError("bad_request", "a 'window' object with a numeric 'id' is required (use list_windows / list_apps)");
+        IntPtr h = new IntPtr(id);
+        if (!IsWindow(h)) throw new ToolError("window_gone", "window " + id + " no longer exists; call list_windows again");
+        return h;
+    }
+
+    // ---- observation ----
+    public Dictionary<string, object> GetWindowState(Dictionary<string, object> p)
+    {
+        IntPtr h = ResolveWindow(p);
+        AssertAppApproved(h);
+        bool includeText = Json.GetBool(p, "include_text", false);
+        bool includeShot = Json.GetBool(p, "include_screenshot", false);
+        long maxElements = Json.GetLong(p, "maxElements", 260);
+        long ttl = Json.GetLong(p, "ttlMs", _snapshotTtlMs);
+
+        uint pid;
+        GetWindowThreadProcessId(h, out pid);
+        Dictionary<string, object> result = new Dictionary<string, object>();
+        result["window"] = WindowDict(h, pid, 0);
+
+        // Codex parity: accessibility is null unless include_text was asked for.
+        result["accessibility"] = includeText ? (object)BuildAccessibility(h, maxElements, (int)ttl) : null;
+        bool blank;
+        result["screenshots"] = includeShot ? (object)CaptureScreenshots(h, out blank) : new List<object>();
+        return result;
+    }
+
+    Dictionary<string, object> BuildAccessibility(IntPtr h, long maxElements, int ttlMs)
+    {
+        Snapshot snap = new Snapshot();
+        snap.WindowId = h.ToInt64();
+        snap.At = DateTime.UtcNow;
+
+        StringBuilder tree = new StringBuilder();
+        int index = 0;
+        AutomationElement root = null;
+        try { root = AutomationElement.FromHandle(h); }
+        catch (Exception ex) { throw new ToolError("uia_unavailable", "UI Automation cannot read this window: " + ex.Message); }
+
+        string focused = null;
+        Walk(root, 0, maxElements, snap, tree, ref index, h, ref focused);
+
+        if (index == 0)
+        {
+            // Electron/Canvas apps may not expose a tree until accessibility is requested;
+            // retry once after a short delay.
+            System.Threading.Thread.Sleep(180);
+            try { root = AutomationElement.FromHandle(h); } catch { }
+            Walk(root, 0, maxElements, snap, tree, ref index, h, ref focused);
+        }
+
+        lock (_sync) { _snapshots[snap.WindowId] = snap; }
+
+        Dictionary<string, object> state = new Dictionary<string, object>();
+        state["tree"] = tree.ToString();
+        state["elementCount"] = (long)index;
+        state["elements"] = snap.Summary;
+
+        // Structured context Codex exposes beside the tree. Each is best-effort:
+        // an app that does not implement the pattern simply omits the field.
+        try
+        {
+            AutomationElement doc = root.FindFirst(TreeScope.Descendants,
+                new PropertyCondition(AutomationElement.ControlTypeProperty, ControlType.Document));
+            if (doc != null)
+            {
+                object dp;
+                if (doc.TryGetCurrentPattern(TextPattern.Pattern, out dp))
+                {
+                    string body = ((TextPattern)dp).DocumentRange.GetText(8000);
+                    if (!string.IsNullOrEmpty(body)) state["document_text"] = body;
+                }
+            }
+            object focusPattern;
+            AutomationElement focusedElement = AutomationElement.FocusedElement;
+            if (focusedElement != null && focusedElement.TryGetCurrentPattern(TextPattern.Pattern, out focusPattern))
+            {
+                System.Windows.Automation.Text.TextPatternRange[] ranges = ((TextPattern)focusPattern).GetSelection();
+                if (ranges != null && ranges.Length > 0)
+                {
+                    string selected = ranges[0].GetText(4000);
+                    if (!string.IsNullOrEmpty(selected)) state["selected_text"] = selected;
+                }
+            }
+            Condition selectable = new PropertyCondition(AutomationElement.IsSelectionItemPatternAvailableProperty, true);
+            AutomationElementCollection selectedItems = root.FindAll(TreeScope.Descendants, selectable);
+            List<object> selectedLines = new List<object>();
+            for (int i = 0; i < selectedItems.Count && selectedLines.Count < 20; i++)
+            {
+                try
+                {
+                    object sp;
+                    if (selectedItems[i].TryGetCurrentPattern(SelectionItemPattern.Pattern, out sp) && ((SelectionItemPattern)sp).Current.IsSelected)
+                        selectedLines.Add(SafeString(selectedItems[i].Current.Name) + " (" + SafeRole(selectedItems[i].Current) + ")");
+                }
+                catch { }
+            }
+            if (selectedLines.Count > 0) state["selected_elements"] = selectedLines;
+        }
+        catch { }
+        state["snapshotAt"] = snap.At.ToString("o");
+        state["ttlMs"] = (long)ttlMs;
+        if (focused != null) state["focused_element"] = focused;
+        return state;
+    }
+
+    void Walk(AutomationElement el, int depth, long maxElements, Snapshot snap, StringBuilder sb, ref int index, IntPtr hwnd, ref string focused)
+    {
+        if (el == null) return;
+        AutomationElementCollection kids;
+        try { kids = el.FindAll(TreeScope.Children, Condition.TrueCondition); }
+        catch { return; }
+
+        for (int i = 0; i < kids.Count; i++)
+        {
+            if (index >= maxElements) return;
+            AutomationElement k;
+            try { k = kids[i]; } catch { continue; }
+            AutomationElement.AutomationElementInformation info;
+            bool offscreen;
+            try { info = k.Current; offscreen = info.IsOffscreen; } catch { continue; }
+            if (offscreen) continue;
+            try
+            {
+
+            long idx = index++;
+            snap.Elements[idx] = k;
+
+            string role = SafeRole(info);
+            string name = SafeString(info.Name);
+            System.Windows.Rect rect = System.Windows.Rect.Empty;
+            try { rect = info.BoundingRectangle; } catch { }
+            bool hasRect = !rect.IsEmpty && rect.Width > 0 && rect.Height > 0;
+
+            Dictionary<string, object> summary = new Dictionary<string, object>();
+            summary["index"] = idx;
+            summary["role"] = role;
+            summary["name"] = name;
+            if (hasRect)
+            {
+                summary["x"] = (int)rect.X;
+                summary["y"] = (int)rect.Y;
+                summary["width"] = (int)rect.Width;
+                summary["height"] = (int)rect.Height;
+            }
+            snap.Summary.Add(summary);
+
+            sb.Append('[').Append(idx).Append("] ").Append(role);
+            if (name.Length > 0) sb.Append(" \"").Append(Trim(name, 120)).Append('"');
+            if (hasRect) sb.Append(" @").Append((int)rect.X).Append(',').Append((int)rect.Y).Append(' ').Append((int)rect.Width).Append('x').Append((int)rect.Height);
+
+            List<string> actions = ActionsOf(k);
+            if (actions.Count > 0) sb.Append(" [").Append(string.Join(",", actions.ToArray())).Append(']');
+
+            StringBuilder val = new StringBuilder();
+            try
+            {
+                object vp;
+                if (k.TryGetCurrentPattern(ValuePattern.Pattern, out vp))
+                {
+                    string v = ((ValuePattern)vp).Current.Value;
+                    if (!string.IsNullOrEmpty(v)) val.Append(" value=\"").Append(Trim(v, 90)).Append('"');
+                }
+                else
+                {
+                    object tp;
+                    if (k.TryGetCurrentPattern(TextPattern.Pattern, out tp))
+                    {
+                        string t = ((TextPattern)tp).DocumentRange.GetText(120);
+                        if (!string.IsNullOrEmpty(t)) val.Append(" text=\"").Append(Trim(t.Replace("\r", " ").Replace("\n", " "), 90)).Append('"');
+                    }
+                }
+            }
+            catch { }
+            sb.Append(val.ToString());
+
+            try { if (info.HasKeyboardFocus && focused == null) focused = "[" + idx + "] " + role + " \"" + name + "\""; }
+            catch { }
+
+            sb.Append('\n');
+            if (depth < 18) Walk(k, depth + 1, maxElements, snap, sb, ref index, hwnd, ref focused);
+            }
+            catch (Exception) { continue; }
+        }
+    }
+
+    static string SafeRole(AutomationElement.AutomationElementInformation info)
+    {
+        try { return info.ControlType.ProgrammaticName.Replace("ControlType.", ""); }
+        catch { return "Unknown"; }
+    }
+
+    static long ToLong(object v) { try { return Convert.ToInt64(v, System.Globalization.CultureInfo.InvariantCulture); } catch { return 0L; } }
+    static double ToDouble(object v) { try { return Convert.ToDouble(v, System.Globalization.CultureInfo.InvariantCulture); } catch { return 1.0; } }
+
+    static string SafeString(string s) { return s == null ? "" : s.Replace("\r", " ").Replace("\n", " ").Trim(); }
+
+    static string Trim(string s, int max) { return s.Length <= max ? s : s.Substring(0, max) + "…"; }
+
+    static List<string> ActionsOf(AutomationElement el)
+    {
+        List<string> a = new List<string>();
+        object pattern;
+        try { if (el.TryGetCurrentPattern(InvokePattern.Pattern, out pattern)) a.Add("Invoke"); } catch { }
+        try { if (el.TryGetCurrentPattern(TogglePattern.Pattern, out pattern)) a.Add("Toggle"); } catch { }
+        try { if (el.TryGetCurrentPattern(SelectionItemPattern.Pattern, out pattern)) a.Add("Select"); } catch { }
+        try { if (el.TryGetCurrentPattern(ExpandCollapsePattern.Pattern, out pattern)) a.Add("ExpandCollapse"); } catch { }
+        try { if (el.TryGetCurrentPattern(ScrollPattern.Pattern, out pattern)) a.Add("Scroll"); } catch { }
+        try { if (el.TryGetCurrentPattern(ValuePattern.Pattern, out pattern)) a.Add("SetValue"); } catch { }
+        return a;
+    }
+
+    /// Every top-level window that belongs to the same "visual state" as the
+    /// target: the window itself plus owned/transient popups (menus, tooltips,
+    /// combo drop-downs) that a screenshot of the window alone would hide.
+    List<IntPtr> RelatedWindows(IntPtr target)
+    {
+        RECT main = BoundsOf(target);
+        uint targetPid;
+        GetWindowThreadProcessId(target, out targetPid);
+        List<IntPtr> result = new List<IntPtr>();
+        List<IntPtr> all = EnumerateWindows(false);
+        for (int i = 0; i < all.Count; i++)
+        {
+            IntPtr h = all[i];
+            if (h == target) continue;
+            if (!IsWindowVisible(h)) continue;
+            if (IsIconic(h)) continue;
+            uint pid;
+            GetWindowThreadProcessId(h, out pid);
+            if (pid != targetPid) continue;
+            bool owned = GetWindow(h, GW_OWNER) == target;
+            bool popup = (GetWindowLong(h, GWL_STYLE) & WS_POPUP) != 0;
+            if (!owned && !popup) continue;
+            RECT r = BoundsOf(h);
+            bool intersects = r.Left < main.Right && r.Right > main.Left && r.Top < main.Bottom && r.Bottom > main.Top;
+            if (!intersects) continue;
+            result.Add(h);
+        }
+        return result;
+    }
+
+    List<object> CaptureScreenshots(IntPtr h, out bool anyBlank)
+    {
+        anyBlank = false;
+        if (IsIconic(h))
+            throw new ToolError("window_minimized", "window " + h.ToInt64() + " is minimized, so there is nothing to observe: call activate_window (or restore it) and observe again. A minimized window must never be captured, because the image would not be the window.");
+        if (!IsWindowVisible(h))
+            throw new ToolError("window_not_visible", "window " + h.ToInt64() + " is not visible; activate it before observing.");
+
+        List<IntPtr> order = EnumerateWindows(true); // z-order, topmost first
+        List<object> shots = new List<object>();
+        List<IntPtr> targets = new List<IntPtr>();
+        targets.Add(h);
+        targets.AddRange(RelatedWindows(h));
+
+        int total = Math.Max(1, order.Count);
+        for (int i = 0; i < targets.Count; i++)
+        {
+            IntPtr w = targets[i];
+            int zIndex = order.IndexOf(w);
+            if (zIndex < 0) zIndex = total - 1;
+            Dictionary<string, object> shot = CaptureWindow(w);
+            shot["zIndex"] = (long)(total - zIndex); // larger = visually above
+            string id = "shot-" + _shotGeneration + "-" + (++_shotSeq).ToString() + "-" + w.ToInt64().ToString();
+            shot["id"] = id;
+            RECT r = BoundsOf(w);
+            lock (_sync)
+            {
+                PruneShots();
+                _shots[id] = new ShotRegion
+                {
+                    WindowId = h.ToInt64(), OriginX = r.Left, OriginY = r.Top,
+                    Width = (int)ToLong(shot["width"]), Height = (int)ToLong(shot["height"]),
+                    ScaleX = ToDouble(shot["scaleX"]), ScaleY = ToDouble(shot["scaleY"]),
+                    At = DateTime.UtcNow,
+                };
+            }
+            shots.Add(shot);
+        }
+        return shots;
+    }
+
+    void PruneShots()
+    {
+        if (_shots.Count <= 48) return;
+        List<string> stale = new List<string>();
+        DateTime cutoff = DateTime.UtcNow.AddMilliseconds(-_shotTtlMs);
+        foreach (KeyValuePair<string, ShotRegion> kv in _shots)
+            if (kv.Value.At < cutoff) stale.Add(kv.Key);
+        for (int i = 0; i < stale.Count; i++) _shots.Remove(stale[i]);
+    }
+
+    ShotRegion ResolveShot(IntPtr h, string screenshotId)
+    {
+        if (string.IsNullOrEmpty(screenshotId))
+            throw new ToolError("bad_request", "screenshotId is empty");
+        ShotRegion region;
+        lock (_sync)
+        {
+            if (!_shots.TryGetValue(screenshotId, out region))
+                throw new ToolError("stale_screenshot", "screenshot '" + screenshotId + "' is no longer cached: it must come from the latest get_window_state for this window. Take a fresh screenshot before retrying.");
+        }
+        if (region.WindowId != h.ToInt64())
+            throw new ToolError("screenshot_mismatch", "screenshot '" + screenshotId + "' belongs to another window; take a fresh observation for this window.");
+        if ((DateTime.UtcNow - region.At).TotalMilliseconds > _shotTtlMs)
+            throw new ToolError("stale_screenshot", "screenshot '" + screenshotId + "' is older than the screenshot TTL. Take a fresh screenshot before retrying.");
+        return region;
+    }
+
+    /// Codex rejects a point that falls outside the screenshot the caller referenced
+    /// ("is outside screenshot bounds"), because that usually means a stale observation.
+    static void AssertPointInsideShot(double x, double y, ShotRegion region, string screenshotId, string what)
+    {
+        if (double.IsNaN(x) || double.IsNaN(y)) return;
+        if (x >= 0 && y >= 0 && x <= region.Width && y <= region.Height) return;
+        throw new ToolError("point_outside_screenshot",
+            what + " (" + x.ToString("0.##") + ", " + y.ToString("0.##") + ") is outside screenshot bounds (0,0)-(" +
+            region.Width + "," + region.Height + ") for '" + screenshotId + "'. Take a fresh screenshot with get_window_state before retrying.");
+    }
+
+    /// Windows.Graphics.Capture through the .NET 8 helper: composed window content,
+    /// occlusion-proof, no DPI halo. Falls back to PrintWindow/screen when absent.
+    static Bitmap TryCaptureWithWgc(IntPtr hwnd)
+    {
+        try
+        {
+            string dir = Path.GetDirectoryName(typeof(WindowsDriver).Assembly.Location) ?? ".";
+            string exe = Path.Combine(dir, "capture", "dsh-cu-capture.exe");
+            if (!File.Exists(exe)) return null;
+            string tmp = Path.Combine(Path.GetTempPath(), "dsh-cu-wgc-" + Guid.NewGuid().ToString("N") + ".png");
+            try
+            {
+                long budget = Budget.RemainingMs();
+                int waitMs = (int)Math.Max(150, Math.Min(1200, budget - 150));
+                ProcessStartInfo psi = new ProcessStartInfo(exe, "--hwnd " + hwnd.ToInt64().ToString() + " --out \"" + tmp + "\" --timeout " + waitMs);
+                psi.UseShellExecute = false;
+                psi.CreateNoWindow = true;
+                psi.RedirectStandardOutput = true;
+                psi.RedirectStandardError = true;
+                using (Process p = Process.Start(psi))
+                {
+                    if (p == null) return null;
+                    if (!p.WaitForExit(waitMs + 900)) { try { p.Kill(); } catch { } return null; }
+                    if (p.ExitCode != 0 || !File.Exists(tmp)) return null;
+                    byte[] bytes = File.ReadAllBytes(tmp);
+                    using (MemoryStream ms = new MemoryStream(bytes)) return new Bitmap(ms);
+                }
+            }
+            finally { try { if (File.Exists(tmp)) File.Delete(tmp); } catch { } }
+        }
+        catch { return null; }
+    }
+
+    Dictionary<string, object> CaptureWindow(IntPtr h)
+    {
+        RECT r = BoundsOf(h);
+        int w = Math.Max(1, r.Right - r.Left);
+        int ht = Math.Max(1, r.Bottom - r.Top);
+        // Cap pathological sizes so a maximised multi-monitor window cannot blow the
+        // attachment byte budget.
+        const int cap = 2560;
+        double scale = 1.0;
+        if (w > cap || ht > cap)
+        {
+            scale = Math.Min((double)cap / w, (double)cap / ht);
+            w = Math.Max(1, (int)(w * scale));
+            ht = Math.Max(1, (int)(ht * scale));
+        }
+
+        Bitmap wgc = TryCaptureWithWgc(h);
+        if (wgc != null)
+        {
+            Dictionary<string, object> wg = FinishShot(h, wgc, r, true, 1.0, 1.0);
+            return wg;
+        }
+
+        Bitmap bmp = new Bitmap(w, ht, PixelFormat.Format24bppRgb);
+        bool printed = false;
+        using (Graphics g = Graphics.FromImage(bmp))
+        {
+            IntPtr hdc = g.GetHdc();
+            try { printed = PrintWindow(h, hdc, 2); }
+            finally { g.ReleaseHdc(hdc); }
+        }
+        if (!printed || IsBlank(bmp))
+        {
+            bmp.Dispose();
+            bmp = CaptureFromScreen(r, w, ht);
+        }
+
+        // DPI-virtualized targets render at their own (usually 96 DPI) size while
+        // occupying a larger physical rect, so PrintWindow leaves a black halo.
+        // Crop to the painted region and remember the physical/screenshot ratio so
+        // coordinate actions can map screenshot pixels back to screen pixels.
+        int fullWidth = bmp.Width, fullHeight = bmp.Height;
+        Rectangle painted = PaintedBounds(bmp);
+        if (painted.Width > 16 && painted.Height > 16 && (painted.Width < bmp.Width || painted.Height < bmp.Height))
+        {
+            Bitmap cropped = bmp.Clone(new Rectangle(0, 0, painted.Right, painted.Bottom), bmp.PixelFormat);
+            bmp.Dispose();
+            bmp = cropped;
+            w = bmp.Width;
+            ht = bmp.Height;
+        }
+        double scaleX = (double)fullWidth / Math.Max(1, bmp.Width);
+        double scaleY = (double)fullHeight / Math.Max(1, bmp.Height);
+
+        byte[] png;
+        using (MemoryStream ms = new MemoryStream())
+        {
+            bmp.Save(ms, ImageFormat.Png);
+            png = ms.ToArray();
+        }
+        bmp.Dispose();
+
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["mediaType"] = "image/png";
+        d["base64"] = Convert.ToBase64String(png);
+        d["width"] = (long)w;
+        d["height"] = (long)ht;
+        d["originX"] = (long)r.Left;
+        d["originY"] = (long)r.Top;
+        d["zIndex"] = 1L;
+        d["printed"] = printed;
+        d["scaleX"] = scaleX;
+        d["scaleY"] = scaleY;
+        return d;
+    }
+
+    static Dictionary<string, object> FinishShot(IntPtr h, Bitmap bmp, RECT r, bool printed, double scaleX, double scaleY)
+    {
+        int w = bmp.Width, ht = bmp.Height;
+        byte[] png;
+        using (MemoryStream ms = new MemoryStream())
+        {
+            bmp.Save(ms, ImageFormat.Png);
+            png = ms.ToArray();
+        }
+        bmp.Dispose();
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["mediaType"] = "image/png";
+        d["base64"] = Convert.ToBase64String(png);
+        d["width"] = (long)w;
+        d["height"] = (long)ht;
+        d["originX"] = (long)r.Left;
+        d["originY"] = (long)r.Top;
+        d["zIndex"] = 1L;
+        d["printed"] = printed;
+        d["scaleX"] = scaleX;
+        d["scaleY"] = scaleY;
+        return d;
+    }
+
+    static Bitmap CaptureFromScreen(RECT r, int w, int ht)
+    {
+        Bitmap bmp = new Bitmap(w, ht, PixelFormat.Format24bppRgb);
+        using (Graphics g = Graphics.FromImage(bmp))
+        {
+            try { g.CopyFromScreen(r.Left, r.Top, 0, 0, new Size(w, ht), CopyPixelOperation.SourceCopy); }
+            catch { }
+        }
+        return bmp;
+    }
+
+    /// Bounding box of everything that is not uniform black, i.e. the region the
+    /// target actually painted. Used to strip the DPI-virtualization halo.
+    static Rectangle PaintedBounds(Bitmap bmp)
+    {
+        int right = 0, bottom = 0;
+        int step = Math.Max(1, Math.Min(bmp.Width, bmp.Height) / 200);
+        for (int y = 0; y < bmp.Height; y += step)
+        {
+            for (int x = 0; x < bmp.Width; x += step)
+            {
+                Color c = bmp.GetPixel(x, y);
+                if (c.R > 8 || c.G > 8 || c.B > 8)
+                {
+                    if (x > right) right = x;
+                    if (y > bottom) bottom = y;
+                }
+            }
+        }
+        return new Rectangle(0, 0, Math.Min(bmp.Width, right + step), Math.Min(bmp.Height, bottom + step));
+    }
+
+    static bool IsBlank(Bitmap bmp)
+    {
+        int samples = 0, dark = 0;
+        int stepX = Math.Max(1, bmp.Width / 12);
+        int stepY = Math.Max(1, bmp.Height / 12);
+        for (int y = 0; y < bmp.Height; y += stepY)
+        {
+            for (int x = 0; x < bmp.Width; x += stepX)
+            {
+                Color c = bmp.GetPixel(x, y);
+                samples++;
+                if (c.R < 6 && c.G < 6 && c.B < 6) dark++;
+            }
+        }
+        return samples > 0 && dark == samples;
+    }
+
+    // ---- element resolution ----
+    AutomationElement ElementAt(IntPtr hwnd, long index)
+    {
+        Snapshot snap;
+        lock (_sync)
+        {
+            if (!_snapshots.TryGetValue(hwnd.ToInt64(), out snap))
+                throw new ToolError("stale_snapshot", "no observation for this window yet: call get_window_state with include_text=true before using element_index");
+        }
+        if ((DateTime.UtcNow - snap.At).TotalMilliseconds > _snapshotTtlMs)
+            throw new ToolError("stale_snapshot", "the element snapshot is older than " + _snapshotTtlMs + "ms: call get_window_state again before acting on element indexes");
+        AutomationElement el;
+        if (!snap.Elements.TryGetValue(index, out el))
+            throw new ToolError("bad_index", "element index " + index + " is not part of the latest observation for this window");
+        return el;
+    }
+
+    static Dictionary<string, object> ScreenshotTarget(IntPtr h, AutomationElement el)
+    {
+        RECT obs = BoundsOf(h);
+        System.Windows.Rect r = el.Current.BoundingRectangle;
+        if (r.IsEmpty || r.Width <= 0 || r.Height <= 0)
+            throw new ToolError("no_geometry", "the element has no bounding rectangle; use x/y coordinates instead");
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["screenX"] = (int)r.X + (int)(r.Width / 2);
+        d["screenY"] = (int)r.Y + (int)(r.Height / 2);
+        d["originX"] = obs.Left;
+        d["originY"] = obs.Top;
+        return d;
+    }
+
+    // ---- input ----
+    static void SendMouse(uint flags) { SendMouseData(flags, 0); }
+
+    static void SendMouseData(uint flags, int data)
+    {
+        INPUT inp = new INPUT();
+        inp.type = INPUT_MOUSE;
+        inp.u.mi.dx = 0; inp.u.mi.dy = 0;
+        inp.u.mi.mouseData = (uint)data;
+        inp.u.mi.dwFlags = flags;
+        INPUT[] arr = new INPUT[] { inp };
+        SendInput(1, arr, Marshal.SizeOf(typeof(INPUT)));
+    }
+
+    static void KeyDown(ushort vk) { SendKey(vk, 0, 0); }
+    static void KeyUp(ushort vk) { SendKey(vk, 0, KEYEVENTF_KEYUP); }
+
+    static void SendKey(ushort vk, ushort scan, uint flags)
+    {
+        INPUT inp = new INPUT();
+        inp.type = INPUT_KEYBOARD;
+        inp.u.ki.wVk = vk;
+        inp.u.ki.wScan = scan;
+        inp.u.ki.dwFlags = flags;
+        INPUT[] arr = new INPUT[] { inp };
+        SendInput(1, arr, Marshal.SizeOf(typeof(INPUT)));
+    }
+
+    static void SendUnicode(char ch, bool up)
+    {
+        INPUT inp = new INPUT();
+        inp.type = INPUT_KEYBOARD;
+        inp.u.ki.wVk = 0;
+        inp.u.ki.wScan = ch;
+        inp.u.ki.dwFlags = KEYEVENTF_UNICODE | (up ? KEYEVENTF_KEYUP : 0);
+        INPUT[] arr = new INPUT[] { inp };
+        SendInput(1, arr, Marshal.SizeOf(typeof(INPUT)));
+    }
+
+    void ClickAt(int screenX, int screenY, int count, string button, Dictionary<string, object> p)
+    {
+        SetCursorPos(screenX, screenY);
+        System.Threading.Thread.Sleep(35);
+        for (int i = 0; i < Math.Max(1, count); i++)
+        {
+            CheckCancel(p);
+            if (i > 0) System.Threading.Thread.Sleep(60);
+            uint down = MOUSEEVENTF_LEFTDOWN, up = MOUSEEVENTF_LEFTUP;
+            if (button == "r") { down = MOUSEEVENTF_RIGHTDOWN; up = MOUSEEVENTF_RIGHTUP; }
+            else if (button == "m") { down = MOUSEEVENTF_MIDDLEDOWN; up = MOUSEEVENTF_MIDDLEUP; }
+            SendMouse(down);
+            try { System.Threading.Thread.Sleep(22); }
+            finally { SendMouse(up); }
+        }
+    }
+
+    public Dictionary<string, object> Click(Dictionary<string, object> p)
+    {
+        CheckCancel(p);
+        AssertUnlocked();
+        EnsureOverlay(8000);
+        IntPtr h = ResolveWindow(p);
+        AssertAppApproved(h);
+        long elementIndex = Json.GetLong(p, "element_index", -1);
+        string button = NormalizeButton(Json.GetString(p, "mouse_button"));
+        int count = (int)Json.GetLong(p, "click_count", 1);
+        int sx, sy;
+
+        if (elementIndex >= 0)
+        {
+            AutomationElement el = ElementAt(h, elementIndex);
+            Dictionary<string, object> t = ScreenshotTarget(h, el);
+            sx = (int)t["screenX"]; sy = (int)t["screenY"];
+            // Prefer the semantic invoke when the element exposes one and we are
+            // doing a plain left single click: it survives occlusion and does not
+            // depend on pixel geometry.
+            if (count == 1 && button == "l")
+            {
+                object pattern;
+                try
+                {
+                    if (el.TryGetCurrentPattern(InvokePattern.Pattern, out pattern))
+                    {
+                        ((InvokePattern)pattern).Invoke();
+                        Dictionary<string, object> r2 = new Dictionary<string, object>();
+                        r2["clicked"] = "invoke"; r2["element_index"] = elementIndex;
+                        return r2;
+                    }
+                }
+                catch (Exception) { }
+            }
+        }
+        else
+        {
+            double x = Json.GetDouble(p, "x", double.NaN);
+            double y = Json.GetDouble(p, "y", double.NaN);
+            if (double.IsNaN(x) || double.IsNaN(y)) throw new ToolError("bad_request", "click needs either 'element_index' or both 'x' and 'y' (window-relative pixels)");
+            string screenshotId = Json.GetString(p, "screenshotId");
+            int originX, originY;
+            double coordScaleX = 1.0, coordScaleY = 1.0;
+            if (!string.IsNullOrEmpty(screenshotId))
+            {
+                ShotRegion region = ResolveShot(h, screenshotId);
+                AssertPointInsideShot(x, y, region, screenshotId, "click");
+                originX = region.OriginX; originY = region.OriginY;
+                coordScaleX = region.ScaleX; coordScaleY = region.ScaleY;
+            }
+            else
+            {
+                RECT bounds = BoundsOf(h);
+                originX = bounds.Left; originY = bounds.Top;
+            }
+            sx = originX + (int)Math.Round(x * coordScaleX);
+            sy = originY + (int)Math.Round(y * coordScaleY);
+        }
+
+        Activate(h); // pointer clicks need a real foreground window
+        MarkCursor(sx, sy);
+        CheckCancel(p);
+        ClickAt(sx, sy, count, button, p);
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["clicked"] = "pointer";
+        r["screenX"] = (long)sx;
+        r["screenY"] = (long)sy;
+        r["clickCount"] = (long)count;
+        return r;
+    }
+
+    static string NormalizeButton(string b)
+    {
+        if (string.IsNullOrEmpty(b)) return "l";
+        string s = b.Trim().ToLowerInvariant();
+        if (s == "right" || s == "r") return "r";
+        if (s == "middle" || s == "m") return "m";
+        return "l";
+    }
+
+    public Dictionary<string, object> PressKey(Dictionary<string, object> p)
+    {
+        CheckCancel(p);
+        AssertUnlocked();
+        EnsureOverlay(8000);
+        IntPtr h = ResolveWindow(p);
+        AssertAppApproved(h);
+        string key = Json.GetString(p, "key");
+        if (string.IsNullOrEmpty(key)) throw new ToolError("bad_request", "press_key requires 'key'");
+        Activate(h);
+        Chord(key, p);
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["pressed"] = key;
+        return r;
+    }
+
+    static readonly Dictionary<string, ushort> KeyMap = BuildKeyMap();
+
+    static Dictionary<string, ushort> BuildKeyMap()
+    {
+        Dictionary<string, ushort> m = new Dictionary<string, ushort>(StringComparer.OrdinalIgnoreCase);
+        for (char c = 'a'; c <= 'z'; c++) m[c.ToString()] = (ushort)(0x41 + (c - 'a'));
+        for (char c = 'A'; c <= 'Z'; c++) m[c.ToString()] = (ushort)(0x41 + (c - 'A'));
+        for (char c = '0'; c <= '9'; c++) m[c.ToString()] = (ushort)(0x30 + (c - '0'));
+        m["return"] = 0x0D; m["enter"] = 0x0D; m["kp_enter"] = 0x0D;
+        m["tab"] = 0x09; m["space"] = 0x20; m["backspace"] = 0x08; m["escape"] = 0x1B; m["esc"] = 0x1B;
+        m["delete"] = 0x2E; m["insert"] = 0x2D; m["home"] = 0x24; m["end"] = 0x23;
+        m["page_up"] = 0x21; m["pageup"] = 0x21; m["prior"] = 0x21; m["page_down"] = 0x22; m["pagedown"] = 0x22; m["next"] = 0x22;
+        m["left"] = 0x25; m["up"] = 0x26; m["right"] = 0x27; m["down"] = 0x28;
+        m["shift_l"] = 0xA0; m["shift"] = 0xA0; m["shift_r"] = 0xA1;
+        m["control_l"] = 0xA2; m["control"] = 0xA2; m["ctrl"] = 0xA2; m["control_r"] = 0xA3;
+        m["alt_l"] = 0xA4; m["alt"] = 0xA4; m["alt_r"] = 0xA5;
+        m["super_l"] = 0x5B; m["win"] = 0x5B; m["meta"] = 0x5B; m["lwin"] = 0x5B; m["super"] = 0x5B;
+        m["super_r"] = 0x5C; m["rwin"] = 0x5C;
+        m["caps_lock"] = 0x14; m["num_lock"] = 0x90; m["scroll_lock"] = 0x91;
+        m["menu"] = 0x5D; m["apps"] = 0x5D; m["print"] = 0x2C; m["pause"] = 0x13;
+        m["period"] = 0xBE; m["greater"] = 0xBE; m["comma"] = 0xBC; m["less"] = 0xBC;
+        m["slash"] = 0xBF; m["question"] = 0xBF; m["backslash"] = 0xDC; m["bar"] = 0xDC;
+        m["semicolon"] = 0xBA; m["colon"] = 0xBA; m["apostrophe"] = 0xDE; m["quote"] = 0xDE;
+        m["minus"] = 0xBD; m["underscore"] = 0xBD; m["equal"] = 0xBB; m["plus"] = 0xBB;
+        m["bracketleft"] = 0xDB; m["bracketright"] = 0xDD; m["grave"] = 0xC0; m["asciitilde"] = 0xC0;
+        for (int i = 1; i <= 24; i++) { m["f" + i] = (ushort)(0x6F + i); }
+        // Numpad: both X keysym style (KP_0) and the aliases the API documents.
+        for (int i = 0; i <= 9; i++)
+        {
+            m["kp_" + i] = (ushort)(0x60 + i);
+            m["numpad_" + i] = (ushort)(0x60 + i);
+        }
+        m["kp_add"] = 0x6B; m["numpad_add"] = 0x6B; m["plus_numpad"] = 0x6B;
+        m["kp_subtract"] = 0x6D; m["numpad_subtract"] = 0x6D; m["minus_numpad"] = 0x6D;
+        m["kp_multiply"] = 0x6A; m["numpad_multiply"] = 0x6A; m["kp_asterisk"] = 0x6A;
+        m["kp_divide"] = 0x6F; m["numpad_divide"] = 0x6F; m["kp_slash"] = 0x6F;
+        m["kp_decimal"] = 0x6E; m["numpad_decimal"] = 0x6E; m["kp_separator"] = 0x6C;
+        m["numpad_enter"] = 0x0D;
+        return m;
+    }
+
+    static readonly HashSet<ushort> ExtendedKeys = new HashSet<ushort>(new ushort[] {
+        0x21,0x22,0x23,0x24,0x25,0x26,0x27,0x28,0x2D,0x2E,0x5B,0x5C,0x5D,0x90,0x6F,0xA3,0xA5
+    });
+
+    static void SendVk(ushort vk, bool up)
+    {
+        uint flags = up ? KEYEVENTF_KEYUP : 0;
+        if (ExtendedKeys.Contains(vk)) flags |= KEYEVENTF_EXTENDEDKEY;
+        SendKey(vk, 0, flags);
+    }
+
+    void Chord(string chord, Dictionary<string, object> p)
+    {
+        string[] parts = chord.Split('+');
+        List<ushort> keys = new List<ushort>();
+        foreach (string raw in parts)
+        {
+            string token = raw.Trim();
+            if (token.Length == 0) continue;
+            keys.Add(ResolveKey(token));
+        }
+        if (keys.Count == 0) throw new ToolError("bad_request", "empty key chord");
+        int pressed = 0;
+        try
+        {
+            for (int i = 0; i < keys.Count; i++)
+            {
+                CheckCancel(p);
+                SendVk(keys[i], false); pressed++;
+                System.Threading.Thread.Sleep(12);
+            }
+        }
+        finally { for (int i = pressed - 1; i >= 0; i--) SendVk(keys[i], true); }
+    }
+
+    /// The Windows/Meta key family is hard-denied: it opens Run, Settings and
+    /// shell surfaces that Computer Use must never drive. Enforced in the helper
+    /// as well as in the plugin, so a direct protocol call cannot bypass it.
+    static readonly HashSet<string> DeniedKeys = new HashSet<string>(StringComparer.OrdinalIgnoreCase) {
+        "win", "win_l", "win_r", "lwin", "rwin", "windows", "windows_l", "windows_r",
+        "super", "super_l", "super_r", "meta", "meta_l", "meta_r", "cmd", "command", "os", "os_l", "os_r"
+    };
+
+    static ushort ResolveKey(string token)
+    {
+        if (DeniedKeys.Contains(token))
+            throw new ToolError("hard_denied", "the Windows/Meta/Super key ('" + token + "') is hard-denied: it opens Run, Settings and shell surfaces that Computer Use must never drive.");
+        ushort vk;
+        if (KeyMap.TryGetValue(token, out vk)) return vk;
+        if (token.Length == 1)
+        {
+            char c = token[0];
+            short scan = VkKeyScan(c);
+            if (scan != -1) return (ushort)(scan & 0xFF);
+            return (ushort)char.ToUpperInvariant(c);
+        }
+        throw new ToolError("bad_key", "unknown key name '" + token + "'");
+    }
+
+    [DllImport("user32.dll")] static extern short VkKeyScan(char ch);
+
+    public Dictionary<string, object> TypeText(Dictionary<string, object> p)
+    {
+        CheckCancel(p);
+        AssertUnlocked();
+        EnsureOverlay(8000);
+        IntPtr h = ResolveWindow(p);
+        AssertAppApproved(h);
+        string text = Json.GetString(p, "text");
+        if (text == null) throw new ToolError("bad_request", "type_text requires 'text'");
+        Activate(h);
+        for (int i = 0; i < text.Length; i++)
+        {
+            CheckCancel(p);
+            char c = text[i];
+            if (c == '\n') { SendVk(0x0D, false); SendVk(0x0D, true); continue; }
+            if (c == '\t') { SendVk(0x09, false); SendVk(0x09, true); continue; }
+            if (c == '\r') continue;
+            SendUnicode(c, false);
+            SendUnicode(c, true);
+            if ((i & 15) == 0) System.Threading.Thread.Sleep(4);
+        }
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["typed"] = (long)text.Length;
+        return r;
+    }
+
+    public Dictionary<string, object> Scroll(Dictionary<string, object> p)
+    {
+        CheckCancel(p);
+        AssertUnlocked();
+        EnsureOverlay(8000);
+        IntPtr h = ResolveWindow(p);
+        AssertAppApproved(h);
+        // Validation first: an invalid call must not activate a window or move the pointer.
+        double scrollY = Json.GetDouble(p, "scrollY", 0);
+        double scrollX = Json.GetDouble(p, "scrollX", 0);
+        double x = Json.GetDouble(p, "x", double.NaN);
+        double y = Json.GetDouble(p, "y", double.NaN);
+        if (double.IsNaN(x) || double.IsNaN(y))
+            throw new ToolError("bad_request", "scroll requires finite 'x' and 'y' (the point to scroll from, in window/screenshot coordinates); click a pane first if it needs focus");
+        if (Math.Abs(scrollX) < 0.0001 && Math.Abs(scrollY) < 0.0001)
+            throw new ToolError("bad_request", "scroll requires a non-zero scrollX or scrollY");
+        Activate(h);
+        int sx, sy;
+        string scrollShot = Json.GetString(p, "screenshotId");
+        int originLeft, originTop;
+        double scrollScaleX = 1.0, scrollScaleY = 1.0;
+        if (!string.IsNullOrEmpty(scrollShot))
+        {
+            ShotRegion region = ResolveShot(h, scrollShot);
+            AssertPointInsideShot(x, y, region, scrollShot, "scroll origin");
+            originLeft = region.OriginX; originTop = region.OriginY;
+            scrollScaleX = region.ScaleX; scrollScaleY = region.ScaleY;
+        }
+        else
+        {
+            RECT b = BoundsOf(h);
+            originLeft = b.Left; originTop = b.Top;
+        }
+        sx = originLeft + (int)Math.Round(x * scrollScaleX);
+        sy = originTop + (int)Math.Round(y * scrollScaleY);
+        MarkCursor(sx, sy);
+        SetCursorPos(sx, sy);
+        System.Threading.Thread.Sleep(40);
+        if (Math.Abs(scrollY) > 0.0001) SendMouseData(MOUSEEVENTF_WHEEL, (int)(-scrollY * 120));
+        if (Math.Abs(scrollX) > 0.0001) SendMouseData(MOUSEEVENTF_HWHEEL, (int)(-scrollX * 120));
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["scrolled"] = true;
+        r["screenX"] = (long)sx; r["screenY"] = (long)sy;
+        return r;
+    }
+
+    public Dictionary<string, object> Drag(Dictionary<string, object> p)
+    {
+        CheckCancel(p);
+        AssertUnlocked();
+        EnsureOverlay(8000);
+        IntPtr h = ResolveWindow(p);
+        AssertAppApproved(h);
+        double fromX = Json.GetDouble(p, "from_x", double.NaN);
+        double fromY = Json.GetDouble(p, "from_y", double.NaN);
+        double toX = Json.GetDouble(p, "to_x", double.NaN);
+        double toY = Json.GetDouble(p, "to_y", double.NaN);
+        if (double.IsNaN(fromX) || double.IsNaN(fromY) || double.IsNaN(toX) || double.IsNaN(toY))
+            throw new ToolError("bad_request", "drag requires finite from_x/from_y/to_x/to_y (window-relative pixels)");
+        Activate(h);
+        int originX, originY;
+        double dragScaleX = 1.0, dragScaleY = 1.0;
+        string dragShot = Json.GetString(p, "screenshotId");
+        if (!string.IsNullOrEmpty(dragShot))
+        {
+            ShotRegion region = ResolveShot(h, dragShot);
+            AssertPointInsideShot(fromX, fromY, region, dragShot, "drag start");
+            AssertPointInsideShot(toX, toY, region, dragShot, "drag end");
+            originX = region.OriginX; originY = region.OriginY;
+            dragScaleX = region.ScaleX; dragScaleY = region.ScaleY;
+        }
+        else
+        {
+            RECT b = BoundsOf(h);
+            originX = b.Left; originY = b.Top;
+        }
+        int fx = originX + (int)Math.Round(fromX * dragScaleX);
+        int fy = originY + (int)Math.Round(fromY * dragScaleY);
+        int tx = originX + (int)Math.Round(toX * dragScaleX);
+        int ty = originY + (int)Math.Round(toY * dragScaleY);
+
+        MarkCursor(fx, fy);
+        SetCursorPos(fx, fy);
+        System.Threading.Thread.Sleep(60);
+        CheckCancel(p);
+        SendMouse(MOUSEEVENTF_LEFTDOWN);
+        try
+        {
+            System.Threading.Thread.Sleep(80);
+            const int steps = 14;
+            for (int i = 1; i <= steps; i++)
+            {
+                CheckCancel(p);
+                int x = fx + (tx - fx) * i / steps;
+                int y = fy + (ty - fy) * i / steps;
+                SetCursorPos(x, y);
+                System.Threading.Thread.Sleep(18);
+            }
+            System.Threading.Thread.Sleep(90);
+            CheckCancel(p);
+        }
+        finally { SendMouse(MOUSEEVENTF_LEFTUP); }
+
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["dragged"] = true;
+        r["from"] = new long[] { fx, fy };
+        r["to"] = new long[] { tx, ty };
+        return r;
+    }
+
+    public Dictionary<string, object> SetValue(Dictionary<string, object> p)
+    {
+        CheckCancel(p);
+        AssertUnlocked();
+        EnsureOverlay(8000);
+        IntPtr h = ResolveWindow(p);
+        AssertAppApproved(h);
+        long elementIndex = Json.GetLong(p, "element_index", -1);
+        if (elementIndex < 0) throw new ToolError("bad_request", "set_value requires 'element_index'");
+        string value = Json.GetString(p, "value");
+        if (value == null) throw new ToolError("bad_request", "set_value requires 'value'");
+        AutomationElement el = ElementAt(h, elementIndex);
+        object pattern;
+        if (!el.TryGetCurrentPattern(ValuePattern.Pattern, out pattern))
+            throw new ToolError("pattern_unavailable", "element [" + elementIndex + "] does not expose the Value pattern; use click + type_text instead");
+        ((ValuePattern)pattern).SetValue(value);
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["valueSet"] = true;
+        r["element_index"] = elementIndex;
+        return r;
+    }
+
+    public Dictionary<string, object> PerformSecondaryAction(Dictionary<string, object> p)
+    {
+        CheckCancel(p);
+        AssertUnlocked();
+        EnsureOverlay(8000);
+        IntPtr h = ResolveWindow(p);
+        AssertAppApproved(h);
+        long elementIndex = Json.GetLong(p, "element_index", -1);
+        string action = Json.GetString(p, "action");
+        if (elementIndex < 0 || string.IsNullOrEmpty(action))
+            throw new ToolError("bad_request", "perform_secondary_action requires 'element_index' and 'action'");
+        AutomationElement el = ElementAt(h, elementIndex);
+        string a = action.Trim().ToLowerInvariant();
+        object pattern;
+
+        if ((a == "raise" || a == "focus") && el.TryGetCurrentPattern(InvokePattern.Pattern, out pattern))
+        { ((InvokePattern)pattern).Invoke(); }
+        else if (a == "raise" || a == "focus") { try { el.SetFocus(); } catch (Exception ex) { throw new ToolError("action_failed", ex.Message); } }
+        else if ((a == "invoke" || a == "press") && el.TryGetCurrentPattern(InvokePattern.Pattern, out pattern))
+        { ((InvokePattern)pattern).Invoke(); }
+        else if ((a == "expand" || a == "collapse") && el.TryGetCurrentPattern(ExpandCollapsePattern.Pattern, out pattern))
+        {
+            ExpandCollapsePattern ec = (ExpandCollapsePattern)pattern;
+            if (a == "expand") ec.Expand(); else ec.Collapse();
+        }
+        else if ((a == "toggle") && el.TryGetCurrentPattern(TogglePattern.Pattern, out pattern))
+        { ((TogglePattern)pattern).Toggle(); }
+        else if ((a == "select" || a == "selectitem") && el.TryGetCurrentPattern(SelectionItemPattern.Pattern, out pattern))
+        { ((SelectionItemPattern)pattern).Select(); }
+        else if (a.StartsWith("scroll") && el.TryGetCurrentPattern(ScrollPattern.Pattern, out pattern))
+        {
+            ScrollPattern sp = (ScrollPattern)pattern;
+            if (a.Contains("up")) sp.ScrollVertical(ScrollAmount.SmallDecrement);
+            else if (a.Contains("down")) sp.ScrollVertical(ScrollAmount.SmallIncrement);
+            else if (a.Contains("left")) sp.ScrollHorizontal(ScrollAmount.SmallDecrement);
+            else if (a.Contains("right")) sp.ScrollHorizontal(ScrollAmount.SmallIncrement);
+            else throw new ToolError("bad_action", "unsupported scroll direction in '" + action + "'");
+        }
+        else
+        {
+            throw new ToolError("action_unavailable", "element [" + elementIndex + "] does not expose action '" + action + "'");
+        }
+
+        Dictionary<string, object> r = new Dictionary<string, object>();
+        r["performed"] = action;
+        r["element_index"] = elementIndex;
+        return r;
+    }
+}