@midscene/computer 1.12.4 → 1.12.5-beta-20260907092148.0

package/dist/es/cli.mjs

@@ -91,6 +91,29 @@ class ComputerInputDriver {
         if (void 0 !== modifiers) lib.keyTap(key, modifiers);
         else lib.keyTap(key);
     }
+    keyToggle(key, state, modifiers) {
+        const lib = this.getLibnutOrThrow('keyToggle');
+        if (void 0 !== modifiers) lib.keyToggle(key, state, modifiers);
+        else lib.keyToggle(key, state);
+    }
+    async keyTapWithExplicitModifiers(key, modifiers, delayMs) {
+        const uniqueModifiers = [
+            ...new Set(modifiers)
+        ];
+        if (0 === uniqueModifiers.length) return void this.keyTap(key);
+        const pressedModifiers = [];
+        try {
+            for (const modifier of uniqueModifiers){
+                this.keyToggle(modifier, 'down');
+                pressedModifiers.push(modifier);
+            }
+            await this.delay(delayMs);
+            this.keyTap(key);
+            await this.delay(delayMs);
+        } finally{
+            for (const modifier of pressedModifiers.reverse())this.releaseKey(modifier);
+        }
+    }
     typeString(text) {
         this.getLibnutOrThrow('typeString').typeString(text);
     }
@@ -163,6 +186,15 @@ class ComputerInputDriver {
             this.options.debug(`Failed to release mouse button ${button}: ${error}`);
         }
     }
+    releaseKey(key) {
+        try {
+            const libnut = this.options.getLibnut();
+            node_assert(libnut, 'libnut not initialized');
+            libnut.keyToggle(key, 'up');
+        } catch (error) {
+            this.options.debug(`Failed to release key ${key}: ${error}`);
+        }
+    }
     rejectPendingInputDelays() {
         const error = this.createDestroyedError('in-flight input');
         for (const waitRef of this.pendingInputDelayWaits){
@@ -523,8 +555,10 @@ async function captureScreenshot(options) {
 }
 const SMOOTH_MOVE_STEPS_TAP = 8;
 const SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_STEPS_DRAG = 20;
 const SMOOTH_MOVE_DELAY_TAP = 8;
 const SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_DELAY_DRAG = 10;
 const MOUSE_MOVE_EFFECT_WAIT = 300;
 const CLICK_SETTLE_DELAY = 50;
 const CLICK_HOLD_DURATION = 100;
@@ -1151,7 +1185,7 @@ Available Displays: ${displays.length > 0 ? displays.map((d)=>d.name).join(', ')
     }
     async healthCheck(displays) {
         console.log('[HealthCheck] Starting health check...');
-        console.log("[HealthCheck] @midscene/computer v1.12.4");
+        console.log("[HealthCheck] @midscene/computer v1.12.5-beta-20260907092148.0");
         console.log('[HealthCheck] Taking screenshot...');
         const screenshotTimeout = 15000;
         let timeoutId;
@@ -1387,6 +1421,8 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
             modifiers,
             driver: this.useAppleScript ? "applescript" : 'libnut'
         });
+        const shortcutDelay = this.options?.keyboardShortcutDelay ?? 0;
+        if (!this.useAppleScript && modifiers.length > 0 && shortcutDelay > 0) return void await this.inputDriver.keyTapWithExplicitModifiers(key, modifiers, shortcutDelay);
         this.inputDriver.sendKey(key, modifiers);
     }
     resolveUntargetedScrollPoint(screenSize) {
@@ -1628,7 +1664,10 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
                     await this.moveDisplayPointer(from, 'Mouse did not reach the drag start target');
                     await this.inputDriver.withMouseButton('left', async ()=>{
                         await this.inputDriver.delay(100);
-                        await this.moveDisplayPointer(to, 'Mouse did not reach the drag end target');
+                        await this.moveDisplayPointer(to, 'Mouse did not reach the drag end target', {
+                            smoothSteps: SMOOTH_MOVE_STEPS_DRAG,
+                            smoothDelay: SMOOTH_MOVE_DELAY_DRAG
+                        });
                         await this.inputDriver.delay(100);
                     });
                 }
@@ -1663,6 +1702,7 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
                 }
             }
         });
+        if (options?.keyboardShortcutDelay !== void 0 && (!Number.isFinite(options.keyboardShortcutDelay) || options.keyboardShortcutDelay < 0)) throw new Error('keyboardShortcutDelay must be a finite non-negative number');
         this.options = options;
         this.displayId = options?.displayId;
         this.useAppleScript = 'darwin' === process.platform && options?.keyboardDriver !== 'libnut';
@@ -2067,10 +2107,10 @@ function rdp_device_define_property(obj, key, value) {
 const device_debug = getDebug('rdp:device');
 const device_SMOOTH_MOVE_STEPS_TAP = 8;
 const device_SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
-const SMOOTH_MOVE_STEPS_DRAG = 12;
+const device_SMOOTH_MOVE_STEPS_DRAG = 12;
 const device_SMOOTH_MOVE_DELAY_TAP = 8;
 const device_SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
-const SMOOTH_MOVE_DELAY_DRAG = 10;
+const device_SMOOTH_MOVE_DELAY_DRAG = 10;
 const device_MOUSE_MOVE_EFFECT_WAIT = 300;
 const device_CLICK_HOLD_DURATION = 50;
 const DRAG_HOLD_DURATION = 100;
@@ -2273,8 +2313,8 @@ class RDPDevice {
                     await this.backend.mouseButton('left', 'down');
                     await sleep(DRAG_HOLD_DURATION);
                     await this.movePointer(Math.round(to.x), Math.round(to.y), {
-                        steps: SMOOTH_MOVE_STEPS_DRAG,
-                        stepDelayMs: SMOOTH_MOVE_DELAY_DRAG
+                        steps: device_SMOOTH_MOVE_STEPS_DRAG,
+                        stepDelayMs: device_SMOOTH_MOVE_DELAY_DRAG
                     });
                     await sleep(DRAG_HOLD_DURATION);
                     await this.backend.mouseButton('left', 'up');
@@ -2365,6 +2405,7 @@ function createLocalComputerDevice(opts) {
         keyboardTypeDelay: opts?.keyboardTypeDelay,
         inputStrategy: opts?.inputStrategy,
         keyboardDriver: opts?.keyboardDriver,
+        keyboardShortcutDelay: opts?.keyboardShortcutDelay,
         headless: opts?.headless,
         xvfbResolution: opts?.xvfbResolution,
         keepXvfbAliveUntilProcessExit: opts?.keepXvfbAliveUntilProcessExit
@@ -2422,6 +2463,7 @@ const computerInitArgShape = {
     headless: z.boolean().optional().describe('Start virtual display via Xvfb (Linux local mode only). Ignored when host is set.'),
     keyboardTypeDelay: z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between keystrokes. In "legacy" mode, positive values enable key-by-key input; zero or omitted uses clipboard input.'),
     inputStrategy: z["enum"](inputStrategies).optional().describe('Text input strategy. "legacy" (default) preserves current Computer behavior, "sequential" sends one Unicode code point at a time, and "bulk" uses one backend input operation. "bulk" requires keyboardTypeDelay to be omitted or set to 0.'),
+    keyboardShortcutDelay: z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between modifier transitions and the main key for local libnut shortcuts. Positive values can improve compatibility with full-screen remote-control clients. Ignored in RDP mode and by the macOS AppleScript driver.'),
     host: z.string().optional().describe('RDP host (FQDN or IP). Set this to switch into RDP mode.'),
     port: z.number().optional().describe('RDP port (default 3389). Requires host.'),
     username: z.string().optional().describe('RDP username. Requires host.'),
@@ -2438,7 +2480,7 @@ const computerInitArgShape = {
 function adaptComputerInitArgs(extracted) {
     if (!extracted || 0 === Object.keys(extracted).length) return;
     if (extracted.host) {
-        const { displayId: _d, headless: _h, ...rdpFields } = extracted;
+        const { displayId: _d, headless: _h, keyboardShortcutDelay: _s, ...rdpFields } = extracted;
         const host = normalizeRdpHost(extracted.host);
         return {
             mode: 'rdp',
@@ -2452,6 +2494,7 @@ function adaptComputerInitArgs(extracted) {
         headless: extracted.headless,
         keyboardTypeDelay: extracted.keyboardTypeDelay,
         inputStrategy: extracted.inputStrategy,
+        keyboardShortcutDelay: extracted.keyboardShortcutDelay,
         ...extractAgentBehaviorInitArgs(extracted) ?? {}
     };
 }
@@ -2503,6 +2546,7 @@ class ComputerMidsceneTools extends BaseMidsceneTools {
         const headless = opts?.mode === 'local' ? opts.headless : void 0;
         const keyboardTypeDelay = opts?.keyboardTypeDelay;
         const inputStrategy = opts?.inputStrategy;
+        const keyboardShortcutDelay = opts?.mode === 'local' ? opts.keyboardShortcutDelay : void 0;
         agent_tools_debug('Creating Computer agent with displayId:', displayId || 'primary');
         const agentOpts = {
             ...displayId ? {
@@ -2517,6 +2561,9 @@ class ComputerMidsceneTools extends BaseMidsceneTools {
             ...void 0 !== inputStrategy ? {
                 inputStrategy
             } : {},
+            ...void 0 !== keyboardShortcutDelay ? {
+                keyboardShortcutDelay
+            } : {},
             ...this.options.keepXvfbAliveUntilProcessExit ? {
                 keepXvfbAliveUntilProcessExit: true
             } : {},
@@ -2601,7 +2648,7 @@ const tools = new ComputerMidsceneTools({
 });
 runToolsCLI(tools, 'midscene-computer', {
     stripPrefix: 'computer_',
-    version: "1.12.4",
+    version: "1.12.5-beta-20260907092148.0",
     extraCommands: createReportCliCommands()
 }).catch((e)=>{
     process.exit(reportCLIError(e));

package/dist/es/index.mjs

@@ -91,6 +91,29 @@ class ComputerInputDriver {
         if (void 0 !== modifiers) lib.keyTap(key, modifiers);
         else lib.keyTap(key);
     }
+    keyToggle(key, state, modifiers) {
+        const lib = this.getLibnutOrThrow('keyToggle');
+        if (void 0 !== modifiers) lib.keyToggle(key, state, modifiers);
+        else lib.keyToggle(key, state);
+    }
+    async keyTapWithExplicitModifiers(key, modifiers, delayMs) {
+        const uniqueModifiers = [
+            ...new Set(modifiers)
+        ];
+        if (0 === uniqueModifiers.length) return void this.keyTap(key);
+        const pressedModifiers = [];
+        try {
+            for (const modifier of uniqueModifiers){
+                this.keyToggle(modifier, 'down');
+                pressedModifiers.push(modifier);
+            }
+            await this.delay(delayMs);
+            this.keyTap(key);
+            await this.delay(delayMs);
+        } finally{
+            for (const modifier of pressedModifiers.reverse())this.releaseKey(modifier);
+        }
+    }
     typeString(text) {
         this.getLibnutOrThrow('typeString').typeString(text);
     }
@@ -163,6 +186,15 @@ class ComputerInputDriver {
             this.options.debug(`Failed to release mouse button ${button}: ${error}`);
         }
     }
+    releaseKey(key) {
+        try {
+            const libnut = this.options.getLibnut();
+            node_assert(libnut, 'libnut not initialized');
+            libnut.keyToggle(key, 'up');
+        } catch (error) {
+            this.options.debug(`Failed to release key ${key}: ${error}`);
+        }
+    }
     rejectPendingInputDelays() {
         const error = this.createDestroyedError('in-flight input');
         for (const waitRef of this.pendingInputDelayWaits){
@@ -523,8 +555,10 @@ async function captureScreenshot(options) {
 }
 const SMOOTH_MOVE_STEPS_TAP = 8;
 const SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_STEPS_DRAG = 20;
 const SMOOTH_MOVE_DELAY_TAP = 8;
 const SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_DELAY_DRAG = 10;
 const MOUSE_MOVE_EFFECT_WAIT = 300;
 const CLICK_SETTLE_DELAY = 50;
 const CLICK_HOLD_DURATION = 100;
@@ -1151,7 +1185,7 @@ Available Displays: ${displays.length > 0 ? displays.map((d)=>d.name).join(', ')
     }
     async healthCheck(displays) {
         console.log('[HealthCheck] Starting health check...');
-        console.log("[HealthCheck] @midscene/computer v1.12.4");
+        console.log("[HealthCheck] @midscene/computer v1.12.5-beta-20260907092148.0");
         console.log('[HealthCheck] Taking screenshot...');
         const screenshotTimeout = 15000;
         let timeoutId;
@@ -1387,6 +1421,8 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
             modifiers,
             driver: this.useAppleScript ? "applescript" : 'libnut'
         });
+        const shortcutDelay = this.options?.keyboardShortcutDelay ?? 0;
+        if (!this.useAppleScript && modifiers.length > 0 && shortcutDelay > 0) return void await this.inputDriver.keyTapWithExplicitModifiers(key, modifiers, shortcutDelay);
         this.inputDriver.sendKey(key, modifiers);
     }
     resolveUntargetedScrollPoint(screenSize) {
@@ -1628,7 +1664,10 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
                     await this.moveDisplayPointer(from, 'Mouse did not reach the drag start target');
                     await this.inputDriver.withMouseButton('left', async ()=>{
                         await this.inputDriver.delay(100);
-                        await this.moveDisplayPointer(to, 'Mouse did not reach the drag end target');
+                        await this.moveDisplayPointer(to, 'Mouse did not reach the drag end target', {
+                            smoothSteps: SMOOTH_MOVE_STEPS_DRAG,
+                            smoothDelay: SMOOTH_MOVE_DELAY_DRAG
+                        });
                         await this.inputDriver.delay(100);
                     });
                 }
@@ -1663,6 +1702,7 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
                 }
             }
         });
+        if (options?.keyboardShortcutDelay !== void 0 && (!Number.isFinite(options.keyboardShortcutDelay) || options.keyboardShortcutDelay < 0)) throw new Error('keyboardShortcutDelay must be a finite non-negative number');
         this.options = options;
         this.displayId = options?.displayId;
         this.useAppleScript = 'darwin' === process.platform && options?.keyboardDriver !== 'libnut';
@@ -2110,10 +2150,10 @@ function rdp_device_define_property(obj, key, value) {
 const device_debug = getDebug('rdp:device');
 const device_SMOOTH_MOVE_STEPS_TAP = 8;
 const device_SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
-const SMOOTH_MOVE_STEPS_DRAG = 12;
+const device_SMOOTH_MOVE_STEPS_DRAG = 12;
 const device_SMOOTH_MOVE_DELAY_TAP = 8;
 const device_SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
-const SMOOTH_MOVE_DELAY_DRAG = 10;
+const device_SMOOTH_MOVE_DELAY_DRAG = 10;
 const device_MOUSE_MOVE_EFFECT_WAIT = 300;
 const device_CLICK_HOLD_DURATION = 50;
 const DRAG_HOLD_DURATION = 100;
@@ -2316,8 +2356,8 @@ class RDPDevice {
                     await this.backend.mouseButton('left', 'down');
                     await sleep(DRAG_HOLD_DURATION);
                     await this.movePointer(Math.round(to.x), Math.round(to.y), {
-                        steps: SMOOTH_MOVE_STEPS_DRAG,
-                        stepDelayMs: SMOOTH_MOVE_DELAY_DRAG
+                        steps: device_SMOOTH_MOVE_STEPS_DRAG,
+                        stepDelayMs: device_SMOOTH_MOVE_DELAY_DRAG
                     });
                     await sleep(DRAG_HOLD_DURATION);
                     await this.backend.mouseButton('left', 'up');
@@ -2408,6 +2448,7 @@ function createLocalComputerDevice(opts) {
         keyboardTypeDelay: opts?.keyboardTypeDelay,
         inputStrategy: opts?.inputStrategy,
         keyboardDriver: opts?.keyboardDriver,
+        keyboardShortcutDelay: opts?.keyboardShortcutDelay,
         headless: opts?.headless,
         xvfbResolution: opts?.xvfbResolution,
         keepXvfbAliveUntilProcessExit: opts?.keepXvfbAliveUntilProcessExit
@@ -2465,6 +2506,7 @@ const computerInitArgShape = {
     headless: z.boolean().optional().describe('Start virtual display via Xvfb (Linux local mode only). Ignored when host is set.'),
     keyboardTypeDelay: z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between keystrokes. In "legacy" mode, positive values enable key-by-key input; zero or omitted uses clipboard input.'),
     inputStrategy: z["enum"](inputStrategies).optional().describe('Text input strategy. "legacy" (default) preserves current Computer behavior, "sequential" sends one Unicode code point at a time, and "bulk" uses one backend input operation. "bulk" requires keyboardTypeDelay to be omitted or set to 0.'),
+    keyboardShortcutDelay: z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between modifier transitions and the main key for local libnut shortcuts. Positive values can improve compatibility with full-screen remote-control clients. Ignored in RDP mode and by the macOS AppleScript driver.'),
     host: z.string().optional().describe('RDP host (FQDN or IP). Set this to switch into RDP mode.'),
     port: z.number().optional().describe('RDP port (default 3389). Requires host.'),
     username: z.string().optional().describe('RDP username. Requires host.'),
@@ -2481,7 +2523,7 @@ const computerInitArgShape = {
 function adaptComputerInitArgs(extracted) {
     if (!extracted || 0 === Object.keys(extracted).length) return;
     if (extracted.host) {
-        const { displayId: _d, headless: _h, ...rdpFields } = extracted;
+        const { displayId: _d, headless: _h, keyboardShortcutDelay: _s, ...rdpFields } = extracted;
         const host = normalizeRdpHost(extracted.host);
         return {
             mode: 'rdp',
@@ -2495,6 +2537,7 @@ function adaptComputerInitArgs(extracted) {
         headless: extracted.headless,
         keyboardTypeDelay: extracted.keyboardTypeDelay,
         inputStrategy: extracted.inputStrategy,
+        keyboardShortcutDelay: extracted.keyboardShortcutDelay,
         ...extractAgentBehaviorInitArgs(extracted) ?? {}
     };
 }
@@ -2546,6 +2589,7 @@ class ComputerMidsceneTools extends BaseMidsceneTools {
         const headless = opts?.mode === 'local' ? opts.headless : void 0;
         const keyboardTypeDelay = opts?.keyboardTypeDelay;
         const inputStrategy = opts?.inputStrategy;
+        const keyboardShortcutDelay = opts?.mode === 'local' ? opts.keyboardShortcutDelay : void 0;
         agent_tools_debug('Creating Computer agent with displayId:', displayId || 'primary');
         const agentOpts = {
             ...displayId ? {
@@ -2560,6 +2604,9 @@ class ComputerMidsceneTools extends BaseMidsceneTools {
             ...void 0 !== inputStrategy ? {
                 inputStrategy
             } : {},
+            ...void 0 !== keyboardShortcutDelay ? {
+                keyboardShortcutDelay
+            } : {},
             ...this.options.keepXvfbAliveUntilProcessExit ? {
                 keepXvfbAliveUntilProcessExit: true
             } : {},
@@ -2640,7 +2687,7 @@ class ComputerMidsceneTools extends BaseMidsceneTools {
     }
 }
 function version() {
-    const currentVersion = "1.12.4";
+    const currentVersion = "1.12.5-beta-20260907092148.0";
     console.log(`@midscene/computer v${currentVersion}`);
     return currentVersion;
 }

package/dist/lib/cli.js

@@ -116,6 +116,29 @@ class ComputerInputDriver {
         if (void 0 !== modifiers) lib.keyTap(key, modifiers);
         else lib.keyTap(key);
     }
+    keyToggle(key, state, modifiers) {
+        const lib = this.getLibnutOrThrow('keyToggle');
+        if (void 0 !== modifiers) lib.keyToggle(key, state, modifiers);
+        else lib.keyToggle(key, state);
+    }
+    async keyTapWithExplicitModifiers(key, modifiers, delayMs) {
+        const uniqueModifiers = [
+            ...new Set(modifiers)
+        ];
+        if (0 === uniqueModifiers.length) return void this.keyTap(key);
+        const pressedModifiers = [];
+        try {
+            for (const modifier of uniqueModifiers){
+                this.keyToggle(modifier, 'down');
+                pressedModifiers.push(modifier);
+            }
+            await this.delay(delayMs);
+            this.keyTap(key);
+            await this.delay(delayMs);
+        } finally{
+            for (const modifier of pressedModifiers.reverse())this.releaseKey(modifier);
+        }
+    }
     typeString(text) {
         this.getLibnutOrThrow('typeString').typeString(text);
     }
@@ -188,6 +211,15 @@ class ComputerInputDriver {
             this.options.debug(`Failed to release mouse button ${button}: ${error}`);
         }
     }
+    releaseKey(key) {
+        try {
+            const libnut = this.options.getLibnut();
+            external_node_assert_default()(libnut, 'libnut not initialized');
+            libnut.keyToggle(key, 'up');
+        } catch (error) {
+            this.options.debug(`Failed to release key ${key}: ${error}`);
+        }
+    }
     rejectPendingInputDelays() {
         const error = this.createDestroyedError('in-flight input');
         for (const waitRef of this.pendingInputDelayWaits){
@@ -549,8 +581,10 @@ async function captureScreenshot(options) {
 }
 const SMOOTH_MOVE_STEPS_TAP = 8;
 const SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_STEPS_DRAG = 20;
 const SMOOTH_MOVE_DELAY_TAP = 8;
 const SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_DELAY_DRAG = 10;
 const MOUSE_MOVE_EFFECT_WAIT = 300;
 const CLICK_SETTLE_DELAY = 50;
 const CLICK_HOLD_DURATION = 100;
@@ -1177,7 +1211,7 @@ Available Displays: ${displays.length > 0 ? displays.map((d)=>d.name).join(', ')
     }
     async healthCheck(displays) {
         console.log('[HealthCheck] Starting health check...');
-        console.log("[HealthCheck] @midscene/computer v1.12.4");
+        console.log("[HealthCheck] @midscene/computer v1.12.5-beta-20260907092148.0");
         console.log('[HealthCheck] Taking screenshot...');
         const screenshotTimeout = 15000;
         let timeoutId;
@@ -1413,6 +1447,8 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
             modifiers,
             driver: this.useAppleScript ? "applescript" : 'libnut'
         });
+        const shortcutDelay = this.options?.keyboardShortcutDelay ?? 0;
+        if (!this.useAppleScript && modifiers.length > 0 && shortcutDelay > 0) return void await this.inputDriver.keyTapWithExplicitModifiers(key, modifiers, shortcutDelay);
         this.inputDriver.sendKey(key, modifiers);
     }
     resolveUntargetedScrollPoint(screenSize) {
@@ -1654,7 +1690,10 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
                     await this.moveDisplayPointer(from, 'Mouse did not reach the drag start target');
                     await this.inputDriver.withMouseButton('left', async ()=>{
                         await this.inputDriver.delay(100);
-                        await this.moveDisplayPointer(to, 'Mouse did not reach the drag end target');
+                        await this.moveDisplayPointer(to, 'Mouse did not reach the drag end target', {
+                            smoothSteps: SMOOTH_MOVE_STEPS_DRAG,
+                            smoothDelay: SMOOTH_MOVE_DELAY_DRAG
+                        });
                         await this.inputDriver.delay(100);
                     });
                 }
@@ -1689,6 +1728,7 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
                 }
             }
         });
+        if (options?.keyboardShortcutDelay !== void 0 && (!Number.isFinite(options.keyboardShortcutDelay) || options.keyboardShortcutDelay < 0)) throw new Error('keyboardShortcutDelay must be a finite non-negative number');
         this.options = options;
         this.displayId = options?.displayId;
         this.useAppleScript = 'darwin' === process.platform && options?.keyboardDriver !== 'libnut';
@@ -2094,10 +2134,10 @@ function rdp_device_define_property(obj, key, value) {
 const device_debug = (0, logger_namespaceObject.getDebug)('rdp:device');
 const device_SMOOTH_MOVE_STEPS_TAP = 8;
 const device_SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
-const SMOOTH_MOVE_STEPS_DRAG = 12;
+const device_SMOOTH_MOVE_STEPS_DRAG = 12;
 const device_SMOOTH_MOVE_DELAY_TAP = 8;
 const device_SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
-const SMOOTH_MOVE_DELAY_DRAG = 10;
+const device_SMOOTH_MOVE_DELAY_DRAG = 10;
 const device_MOUSE_MOVE_EFFECT_WAIT = 300;
 const device_CLICK_HOLD_DURATION = 50;
 const DRAG_HOLD_DURATION = 100;
@@ -2300,8 +2340,8 @@ class RDPDevice {
                     await this.backend.mouseButton('left', 'down');
                     await (0, utils_namespaceObject.sleep)(DRAG_HOLD_DURATION);
                     await this.movePointer(Math.round(to.x), Math.round(to.y), {
-                        steps: SMOOTH_MOVE_STEPS_DRAG,
-                        stepDelayMs: SMOOTH_MOVE_DELAY_DRAG
+                        steps: device_SMOOTH_MOVE_STEPS_DRAG,
+                        stepDelayMs: device_SMOOTH_MOVE_DELAY_DRAG
                     });
                     await (0, utils_namespaceObject.sleep)(DRAG_HOLD_DURATION);
                     await this.backend.mouseButton('left', 'up');
@@ -2392,6 +2432,7 @@ function createLocalComputerDevice(opts) {
         keyboardTypeDelay: opts?.keyboardTypeDelay,
         inputStrategy: opts?.inputStrategy,
         keyboardDriver: opts?.keyboardDriver,
+        keyboardShortcutDelay: opts?.keyboardShortcutDelay,
         headless: opts?.headless,
         xvfbResolution: opts?.xvfbResolution,
         keepXvfbAliveUntilProcessExit: opts?.keepXvfbAliveUntilProcessExit
@@ -2449,6 +2490,7 @@ const computerInitArgShape = {
     headless: core_namespaceObject.z.boolean().optional().describe('Start virtual display via Xvfb (Linux local mode only). Ignored when host is set.'),
     keyboardTypeDelay: core_namespaceObject.z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between keystrokes. In "legacy" mode, positive values enable key-by-key input; zero or omitted uses clipboard input.'),
     inputStrategy: core_namespaceObject.z["enum"](device_namespaceObject.inputStrategies).optional().describe('Text input strategy. "legacy" (default) preserves current Computer behavior, "sequential" sends one Unicode code point at a time, and "bulk" uses one backend input operation. "bulk" requires keyboardTypeDelay to be omitted or set to 0.'),
+    keyboardShortcutDelay: core_namespaceObject.z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between modifier transitions and the main key for local libnut shortcuts. Positive values can improve compatibility with full-screen remote-control clients. Ignored in RDP mode and by the macOS AppleScript driver.'),
     host: core_namespaceObject.z.string().optional().describe('RDP host (FQDN or IP). Set this to switch into RDP mode.'),
     port: core_namespaceObject.z.number().optional().describe('RDP port (default 3389). Requires host.'),
     username: core_namespaceObject.z.string().optional().describe('RDP username. Requires host.'),
@@ -2465,7 +2507,7 @@ const computerInitArgShape = {
 function adaptComputerInitArgs(extracted) {
     if (!extracted || 0 === Object.keys(extracted).length) return;
     if (extracted.host) {
-        const { displayId: _d, headless: _h, ...rdpFields } = extracted;
+        const { displayId: _d, headless: _h, keyboardShortcutDelay: _s, ...rdpFields } = extracted;
         const host = normalizeRdpHost(extracted.host);
         return {
             mode: 'rdp',
@@ -2479,6 +2521,7 @@ function adaptComputerInitArgs(extracted) {
         headless: extracted.headless,
         keyboardTypeDelay: extracted.keyboardTypeDelay,
         inputStrategy: extracted.inputStrategy,
+        keyboardShortcutDelay: extracted.keyboardShortcutDelay,
         ...(0, agent_behavior_init_args_namespaceObject.extractAgentBehaviorInitArgs)(extracted) ?? {}
     };
 }
@@ -2530,6 +2573,7 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
         const headless = opts?.mode === 'local' ? opts.headless : void 0;
         const keyboardTypeDelay = opts?.keyboardTypeDelay;
         const inputStrategy = opts?.inputStrategy;
+        const keyboardShortcutDelay = opts?.mode === 'local' ? opts.keyboardShortcutDelay : void 0;
         agent_tools_debug('Creating Computer agent with displayId:', displayId || 'primary');
         const agentOpts = {
             ...displayId ? {
@@ -2544,6 +2588,9 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
             ...void 0 !== inputStrategy ? {
                 inputStrategy
             } : {},
+            ...void 0 !== keyboardShortcutDelay ? {
+                keyboardShortcutDelay
+            } : {},
             ...this.options.keepXvfbAliveUntilProcessExit ? {
                 keepXvfbAliveUntilProcessExit: true
             } : {},
@@ -2628,7 +2675,7 @@ const tools = new ComputerMidsceneTools({
 });
 (0, cli_namespaceObject.runToolsCLI)(tools, 'midscene-computer', {
     stripPrefix: 'computer_',
-    version: "1.12.4",
+    version: "1.12.5-beta-20260907092148.0",
     extraCommands: (0, core_namespaceObject.createReportCliCommands)()
 }).catch((e)=>{
     process.exit((0, cli_namespaceObject.reportCLIError)(e));

package/dist/lib/index.js

@@ -142,6 +142,29 @@ class ComputerInputDriver {
         if (void 0 !== modifiers) lib.keyTap(key, modifiers);
         else lib.keyTap(key);
     }
+    keyToggle(key, state, modifiers) {
+        const lib = this.getLibnutOrThrow('keyToggle');
+        if (void 0 !== modifiers) lib.keyToggle(key, state, modifiers);
+        else lib.keyToggle(key, state);
+    }
+    async keyTapWithExplicitModifiers(key, modifiers, delayMs) {
+        const uniqueModifiers = [
+            ...new Set(modifiers)
+        ];
+        if (0 === uniqueModifiers.length) return void this.keyTap(key);
+        const pressedModifiers = [];
+        try {
+            for (const modifier of uniqueModifiers){
+                this.keyToggle(modifier, 'down');
+                pressedModifiers.push(modifier);
+            }
+            await this.delay(delayMs);
+            this.keyTap(key);
+            await this.delay(delayMs);
+        } finally{
+            for (const modifier of pressedModifiers.reverse())this.releaseKey(modifier);
+        }
+    }
     typeString(text) {
         this.getLibnutOrThrow('typeString').typeString(text);
     }
@@ -214,6 +237,15 @@ class ComputerInputDriver {
             this.options.debug(`Failed to release mouse button ${button}: ${error}`);
         }
     }
+    releaseKey(key) {
+        try {
+            const libnut = this.options.getLibnut();
+            external_node_assert_default()(libnut, 'libnut not initialized');
+            libnut.keyToggle(key, 'up');
+        } catch (error) {
+            this.options.debug(`Failed to release key ${key}: ${error}`);
+        }
+    }
     rejectPendingInputDelays() {
         const error = this.createDestroyedError('in-flight input');
         for (const waitRef of this.pendingInputDelayWaits){
@@ -575,8 +607,10 @@ async function captureScreenshot(options) {
 }
 const SMOOTH_MOVE_STEPS_TAP = 8;
 const SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_STEPS_DRAG = 20;
 const SMOOTH_MOVE_DELAY_TAP = 8;
 const SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
+const SMOOTH_MOVE_DELAY_DRAG = 10;
 const MOUSE_MOVE_EFFECT_WAIT = 300;
 const CLICK_SETTLE_DELAY = 50;
 const CLICK_HOLD_DURATION = 100;
@@ -1203,7 +1237,7 @@ Available Displays: ${displays.length > 0 ? displays.map((d)=>d.name).join(', ')
     }
     async healthCheck(displays) {
         console.log('[HealthCheck] Starting health check...');
-        console.log("[HealthCheck] @midscene/computer v1.12.4");
+        console.log("[HealthCheck] @midscene/computer v1.12.5-beta-20260907092148.0");
         console.log('[HealthCheck] Taking screenshot...');
         const screenshotTimeout = 15000;
         let timeoutId;
@@ -1439,6 +1473,8 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
             modifiers,
             driver: this.useAppleScript ? "applescript" : 'libnut'
         });
+        const shortcutDelay = this.options?.keyboardShortcutDelay ?? 0;
+        if (!this.useAppleScript && modifiers.length > 0 && shortcutDelay > 0) return void await this.inputDriver.keyTapWithExplicitModifiers(key, modifiers, shortcutDelay);
         this.inputDriver.sendKey(key, modifiers);
     }
     resolveUntargetedScrollPoint(screenSize) {
@@ -1680,7 +1716,10 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
                     await this.moveDisplayPointer(from, 'Mouse did not reach the drag start target');
                     await this.inputDriver.withMouseButton('left', async ()=>{
                         await this.inputDriver.delay(100);
-                        await this.moveDisplayPointer(to, 'Mouse did not reach the drag end target');
+                        await this.moveDisplayPointer(to, 'Mouse did not reach the drag end target', {
+                            smoothSteps: SMOOTH_MOVE_STEPS_DRAG,
+                            smoothDelay: SMOOTH_MOVE_DELAY_DRAG
+                        });
                         await this.inputDriver.delay(100);
                     });
                 }
@@ -1715,6 +1754,7 @@ $g.Dispose(); $bmp.Dispose(); $ms.Dispose()
                 }
             }
         });
+        if (options?.keyboardShortcutDelay !== void 0 && (!Number.isFinite(options.keyboardShortcutDelay) || options.keyboardShortcutDelay < 0)) throw new Error('keyboardShortcutDelay must be a finite non-negative number');
         this.options = options;
         this.displayId = options?.displayId;
         this.useAppleScript = 'darwin' === process.platform && options?.keyboardDriver !== 'libnut';
@@ -2164,10 +2204,10 @@ function rdp_device_define_property(obj, key, value) {
 const device_debug = (0, logger_namespaceObject.getDebug)('rdp:device');
 const device_SMOOTH_MOVE_STEPS_TAP = 8;
 const device_SMOOTH_MOVE_STEPS_MOUSE_MOVE = 10;
-const SMOOTH_MOVE_STEPS_DRAG = 12;
+const device_SMOOTH_MOVE_STEPS_DRAG = 12;
 const device_SMOOTH_MOVE_DELAY_TAP = 8;
 const device_SMOOTH_MOVE_DELAY_MOUSE_MOVE = 10;
-const SMOOTH_MOVE_DELAY_DRAG = 10;
+const device_SMOOTH_MOVE_DELAY_DRAG = 10;
 const device_MOUSE_MOVE_EFFECT_WAIT = 300;
 const device_CLICK_HOLD_DURATION = 50;
 const DRAG_HOLD_DURATION = 100;
@@ -2370,8 +2410,8 @@ class RDPDevice {
                     await this.backend.mouseButton('left', 'down');
                     await (0, utils_namespaceObject.sleep)(DRAG_HOLD_DURATION);
                     await this.movePointer(Math.round(to.x), Math.round(to.y), {
-                        steps: SMOOTH_MOVE_STEPS_DRAG,
-                        stepDelayMs: SMOOTH_MOVE_DELAY_DRAG
+                        steps: device_SMOOTH_MOVE_STEPS_DRAG,
+                        stepDelayMs: device_SMOOTH_MOVE_DELAY_DRAG
                     });
                     await (0, utils_namespaceObject.sleep)(DRAG_HOLD_DURATION);
                     await this.backend.mouseButton('left', 'up');
@@ -2462,6 +2502,7 @@ function createLocalComputerDevice(opts) {
         keyboardTypeDelay: opts?.keyboardTypeDelay,
         inputStrategy: opts?.inputStrategy,
         keyboardDriver: opts?.keyboardDriver,
+        keyboardShortcutDelay: opts?.keyboardShortcutDelay,
         headless: opts?.headless,
         xvfbResolution: opts?.xvfbResolution,
         keepXvfbAliveUntilProcessExit: opts?.keepXvfbAliveUntilProcessExit
@@ -2522,6 +2563,7 @@ const computerInitArgShape = {
     headless: core_namespaceObject.z.boolean().optional().describe('Start virtual display via Xvfb (Linux local mode only). Ignored when host is set.'),
     keyboardTypeDelay: core_namespaceObject.z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between keystrokes. In "legacy" mode, positive values enable key-by-key input; zero or omitted uses clipboard input.'),
     inputStrategy: core_namespaceObject.z["enum"](device_namespaceObject.inputStrategies).optional().describe('Text input strategy. "legacy" (default) preserves current Computer behavior, "sequential" sends one Unicode code point at a time, and "bulk" uses one backend input operation. "bulk" requires keyboardTypeDelay to be omitted or set to 0.'),
+    keyboardShortcutDelay: core_namespaceObject.z.number().finite().nonnegative().optional().describe('Finite non-negative delay in milliseconds between modifier transitions and the main key for local libnut shortcuts. Positive values can improve compatibility with full-screen remote-control clients. Ignored in RDP mode and by the macOS AppleScript driver.'),
     host: core_namespaceObject.z.string().optional().describe('RDP host (FQDN or IP). Set this to switch into RDP mode.'),
     port: core_namespaceObject.z.number().optional().describe('RDP port (default 3389). Requires host.'),
     username: core_namespaceObject.z.string().optional().describe('RDP username. Requires host.'),
@@ -2538,7 +2580,7 @@ const computerInitArgShape = {
 function adaptComputerInitArgs(extracted) {
     if (!extracted || 0 === Object.keys(extracted).length) return;
     if (extracted.host) {
-        const { displayId: _d, headless: _h, ...rdpFields } = extracted;
+        const { displayId: _d, headless: _h, keyboardShortcutDelay: _s, ...rdpFields } = extracted;
         const host = normalizeRdpHost(extracted.host);
         return {
             mode: 'rdp',
@@ -2552,6 +2594,7 @@ function adaptComputerInitArgs(extracted) {
         headless: extracted.headless,
         keyboardTypeDelay: extracted.keyboardTypeDelay,
         inputStrategy: extracted.inputStrategy,
+        keyboardShortcutDelay: extracted.keyboardShortcutDelay,
         ...(0, agent_behavior_init_args_namespaceObject.extractAgentBehaviorInitArgs)(extracted) ?? {}
     };
 }
@@ -2603,6 +2646,7 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
         const headless = opts?.mode === 'local' ? opts.headless : void 0;
         const keyboardTypeDelay = opts?.keyboardTypeDelay;
         const inputStrategy = opts?.inputStrategy;
+        const keyboardShortcutDelay = opts?.mode === 'local' ? opts.keyboardShortcutDelay : void 0;
         agent_tools_debug('Creating Computer agent with displayId:', displayId || 'primary');
         const agentOpts = {
             ...displayId ? {
@@ -2617,6 +2661,9 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
             ...void 0 !== inputStrategy ? {
                 inputStrategy
             } : {},
+            ...void 0 !== keyboardShortcutDelay ? {
+                keyboardShortcutDelay
+            } : {},
             ...this.options.keepXvfbAliveUntilProcessExit ? {
                 keepXvfbAliveUntilProcessExit: true
             } : {},
@@ -2698,7 +2745,7 @@ class ComputerMidsceneTools extends base_tools_namespaceObject.BaseMidsceneTools
 }
 const env_namespaceObject = require("@midscene/shared/env");
 function version() {
-    const currentVersion = "1.12.4";
+    const currentVersion = "1.12.5-beta-20260907092148.0";
     console.log(`@midscene/computer v${currentVersion}`);
     return currentVersion;
 }

package/dist/types/index.d.ts

@@ -177,6 +177,17 @@ export declare interface ComputerDeviceOpt extends ComputerDeviceInputOpt {
      * - 'libnut': Use libnut's keyTap (faster but may not work with some TUI apps)
      */
     keyboardDriver?: 'applescript' | 'libnut';
+    /**
+     * Delay in milliseconds between explicit modifier transitions and the main
+     * key for local libnut shortcuts. A positive value changes modified
+     * shortcuts from one `keyTap` call into modifier-down, main-key, and
+     * modifier-up phases. This can help foreground clients whose full-screen
+     * keyboard capture misses rapidly synthesized modifier state changes.
+     *
+     * Ignored by the macOS AppleScript driver and RDP mode.
+     * @default 0
+     */
+    keyboardShortcutDelay?: number;
     /**
      * Headless mode via Xvfb (Linux only).
      * - true: start Xvfb virtual display
@@ -210,7 +221,7 @@ export declare type ComputerInterface = ComputerDevice | RDPDevice;
 /** Init args for the local desktop agent (macOS/Windows/Linux). */
 declare type ComputerLocalInitArgs = {
     mode: 'local';
-} & Pick<ComputerDeviceOpt, 'displayId' | 'headless'> & Pick<ComputerDeviceOpt, 'inputStrategy' | 'keyboardTypeDelay'> & AgentBehaviorInitArgs;
+} & Pick<ComputerDeviceOpt, 'displayId' | 'headless'> & Pick<ComputerDeviceOpt, 'inputStrategy' | 'keyboardTypeDelay' | 'keyboardShortcutDelay'> & AgentBehaviorInitArgs;
 
 /**
  * Computer-specific tools manager

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@midscene/computer",
-  "version": "1.12.4",
+  "version": "1.12.5-beta-20260907092148.0",
   "description": "Midscene.js Computer Desktop Automation",
   "license": "MIT",
   "repository": {
@@ -33,8 +33,8 @@
     "@computer-use/libnut": "^4.2.0",
     "clipboardy": "^4.0.0",
     "screenshot-desktop": "^1.15.3",
-    "@midscene/core": "1.12.4",
-    "@midscene/shared": "1.12.4"
+    "@midscene/core": "1.12.5-beta-20260907092148.0",
+    "@midscene/shared": "1.12.5-beta-20260907092148.0"
   },
   "optionalDependencies": {
     "node-mac-permissions": "2.5.0"