execbro 2.12.3 → 2.12.4

package/build/tools/screenshotTools.js

@@ -5,6 +5,11 @@ import { recordScreenMetrics } from "../core/projectMemory.js";
 import { iosScreenshot, androidScreenshot, detectLogBox, formatLogBoxWarning, getScreenState, formatScreenStateSummary, imageBuffer, getActiveOrBootedSimulatorUdid, connectedApps, getConnectedAppBySimulatorUdid, getConnectedAppByAndroidDeviceId, iosDescribeAll, detectIOSSystemOverlay, formatIOSSystemOverlayWarning, androidGetStatusBarHeight, androidGetDensity, } from "../core/index.js";
 import { screenStateToScreenSpace } from "../core/screenSpace.js";
 import { resolvePhysicalIosDevice, physicalIosScreenshot } from "../core/iosPhysical.js";
+import { writeFileSync } from "node:fs";
+import { chromiumAppFor } from "../core/connection.js";
+import { chromiumScreenState } from "../core/chromiumScreen.js";
+import { resolveDeviceTarget, formatResolverError } from "../core/deviceResolver.js";
+import { chromiumCapture } from "../core/chromium.js";
 async function capturePhysicalDevice(phys, outputPath) {
     const result = await physicalIosScreenshot(phys.udid, outputPath);
     if (!result.success || !result.data) {
@@ -43,243 +48,461 @@ async function capturePhysicalDevice(phys, outputPath) {
         ]
     };
 }
-export function registerScreenshotTools(server) {
-    registerToolWithTelemetry(server, "ios_screenshot", {
-        description: "Take a screenshot from an iOS simulator. Returns the image plus a screen-state summary: active route (name + navigation stack), overlay-grouped tappable elements (pressables behind an open sheet/modal are excluded), component names as JSX tags, labels, testIDs, and frames — all in ready-to-tap pixel coordinates. Prefer tap(text=\"...\") when text is exact and unique; otherwise use tap(x, y) with coordinates from the list — this is the most reliable way to tap icons or visually-identified elements. Use component names for inspect_component/find_components.\n" +
-            "PURPOSE: Snapshot what the user sees on iOS AND get tap-ready pressables + a structured component map in one call.\n" +
-            "WHEN TO USE: Any visual verification, before/after comparison, or as the starting point for tapping UI by coordinates.\n" +
-            "WORKFLOW: ios_screenshot -> pick element from pressables -> tap(x, y) or tap(testID=...) -> ios_screenshot to verify.\n" +
-            "PHYSICAL DEVICES: a USB-attached iPhone/iPad works too (pass its UDID or name from list_devices) — capture only, via pymobiledevice3. No pressables, no tap/swipe/input_text: iOS has no touch injection below 17.\n" +
-            "LIMITATIONS: Otherwise requires a booted iOS simulator (simctl). For system dialogs without RN, combine with tap(..., native=true).\n" +
-            "GOOD: ios_screenshot()\n" +
-            "BAD: ios_screenshot({ udid: \"guess\" }) with a made-up UDID — run list_devices first.\n" +
-            "SOURCE: to jump from a pixel to the code that renders it, call inspect_at_point(x, y) — it returns the absolute file and line.\n",
-        inputSchema: {
-            outputPath: z
-                .string()
-                .optional()
-                .describe("Optional path to save the screenshot. If not provided, saves to temp directory."),
-            udid: z
-                .string()
-                .optional()
-                .describe(IOS_ARG_DESC),
-            device: z
-                .string()
-                .optional()
-                .describe("Alias for `udid` — same accepted values. Provided for consistency with tap/get_screen_layout/get_screen_state, which all use `device`. If both are given, `udid` wins.")
-        }
-    }, async ({ outputPath, udid, device }) => {
-        const resolved = await resolveIosUdid(udid ?? device);
-        if (!resolved.ok) {
-            const hint = udid ?? device;
-            const phys = hint ? await resolvePhysicalIosDevice(hint) : null;
-            if (phys)
-                return await capturePhysicalDevice(phys, outputPath);
-            return resolved.response;
-        }
-        const targetUdid = resolved.udid ?? (await getActiveOrBootedSimulatorUdid());
-        const capturePromise = iosScreenshot(outputPath, targetUdid ?? undefined);
-        const describePromise = iosDescribeAll(targetUdid ?? undefined).catch(() => null);
-        const earlyTargetApp = targetUdid ? getConnectedAppBySimulatorUdid(targetUdid) : null;
-        const screenStatePromise = earlyTargetApp
-            ? getScreenState({ device: earlyTargetApp.deviceInfo.deviceName }).catch(() => null)
-            : Promise.resolve(null);
-        const result = await capturePromise;
-        if (!result.success) {
-            return {
-                content: [
-                    {
-                        type: "text",
-                        text: `Error: ${result.error}`
-                    }
-                ],
-                isError: true
+async function iosScreenshotHandler({ outputPath, udid, device }) {
+    const resolved = await resolveIosUdid(udid ?? device);
+    if (!resolved.ok) {
+        const hint = udid ?? device;
+        const phys = hint ? await resolvePhysicalIosDevice(hint) : null;
+        if (phys)
+            return await capturePhysicalDevice(phys, outputPath);
+        return resolved.response;
+    }
+    const targetUdid = resolved.udid ?? (await getActiveOrBootedSimulatorUdid());
+    const capturePromise = iosScreenshot(outputPath, targetUdid ?? undefined);
+    const describePromise = iosDescribeAll(targetUdid ?? undefined).catch(() => null);
+    const earlyTargetApp = targetUdid ? getConnectedAppBySimulatorUdid(targetUdid) : null;
+    const screenStatePromise = earlyTargetApp
+        ? getScreenState({ device: earlyTargetApp.deviceInfo.deviceName }).catch(() => null)
+        : Promise.resolve(null);
+    const result = await capturePromise;
+    if (!result.success) {
+        return {
+            content: [
+                {
+                    type: "text",
+                    text: `Error: ${result.error}`
+                }
+            ],
+            isError: true
+        };
+    }
+    if (result.data) {
+        const pixelWidth = result.originalWidth || 0;
+        const pixelHeight = result.originalHeight || 0;
+        const resolvedUdid = targetUdid;
+        const targetApp = resolvedUdid ? getConnectedAppBySimulatorUdid(resolvedUdid) : null;
+        const targetDeviceName = targetApp?.deviceInfo.deviceName;
+        if (targetApp) {
+            targetApp.lastScreenshot = {
+                originalWidth: pixelWidth,
+                originalHeight: pixelHeight,
+                scaleFactor: result.scaleFactor || 1,
             };
         }
-        if (result.data) {
-            const pixelWidth = result.originalWidth || 0;
-            const pixelHeight = result.originalHeight || 0;
-            const resolvedUdid = targetUdid;
-            const targetApp = resolvedUdid ? getConnectedAppBySimulatorUdid(resolvedUdid) : null;
-            const targetDeviceName = targetApp?.deviceInfo.deviceName;
-            if (targetApp) {
-                targetApp.lastScreenshot = {
-                    originalWidth: pixelWidth,
-                    originalHeight: pixelHeight,
-                    scaleFactor: result.scaleFactor || 1,
-                };
-            }
-            let pointWidth = 0;
-            let pointHeight = 0;
-            let scaleFactor = 3;
-            let safeAreaTop = 59;
-            try {
-                const describeResult = await describePromise;
-                if (describeResult && describeResult.success && describeResult.elements && describeResult.elements.length > 0) {
-                    const rootElement = describeResult.elements[0];
-                    if (rootElement.frame) {
-                        pointWidth = Math.round(rootElement.frame.width);
-                        pointHeight = Math.round(rootElement.frame.height);
-                        if (rootElement.frame.y > 0) {
-                            safeAreaTop = Math.round(rootElement.frame.y);
-                        }
+        let pointWidth = 0;
+        let pointHeight = 0;
+        let scaleFactor = 3;
+        let safeAreaTop = 59;
+        try {
+            const describeResult = await describePromise;
+            if (describeResult && describeResult.success && describeResult.elements && describeResult.elements.length > 0) {
+                const rootElement = describeResult.elements[0];
+                if (rootElement.frame) {
+                    pointWidth = Math.round(rootElement.frame.width);
+                    pointHeight = Math.round(rootElement.frame.height);
+                    if (rootElement.frame.y > 0) {
+                        safeAreaTop = Math.round(rootElement.frame.y);
                     }
-                    else if (rootElement.AXFrame) {
-                        const match = rootElement.AXFrame.match(/\{\{([\d.]+),\s*([\d.]+)\},\s*\{([\d.]+),\s*([\d.]+)\}\}/);
-                        if (match) {
-                            const frameY = parseFloat(match[2]);
-                            pointWidth = Math.round(parseFloat(match[3]));
-                            pointHeight = Math.round(parseFloat(match[4]));
-                            if (frameY > 0) {
-                                safeAreaTop = Math.round(frameY);
-                            }
+                }
+                else if (rootElement.AXFrame) {
+                    const match = rootElement.AXFrame.match(/\{\{([\d.]+),\s*([\d.]+)\},\s*\{([\d.]+),\s*([\d.]+)\}\}/);
+                    if (match) {
+                        const frameY = parseFloat(match[2]);
+                        pointWidth = Math.round(parseFloat(match[3]));
+                        pointHeight = Math.round(parseFloat(match[4]));
+                        if (frameY > 0) {
+                            safeAreaTop = Math.round(frameY);
                         }
                     }
-                    if (pointWidth > 0) {
-                        scaleFactor = Math.round(pixelWidth / pointWidth);
-                    }
+                }
+                if (pointWidth > 0) {
+                    scaleFactor = Math.round(pixelWidth / pointWidth);
+                }
+            }
+        }
+        catch {
+        }
+        if (pointWidth === 0) {
+            pointWidth = Math.round(pixelWidth / scaleFactor);
+            pointHeight = Math.round(pixelHeight / scaleFactor);
+        }
+        const safeAreaOffsetPixels = safeAreaTop * scaleFactor;
+        let pressablesText = null;
+        let pressablesIsScreenState = false;
+        if (targetApp) {
+            try {
+                const ssResult = await screenStatePromise;
+                if (ssResult && ssResult.success && ssResult.screenState) {
+                    const screenshotScale = result.scaleFactor || 1;
+                    const toPx = (v) => Math.round((v * scaleFactor) / screenshotScale);
+                    const screenSpaceSs = screenStateToScreenSpace(ssResult.screenState, {
+                        platform: "ios",
+                        topInset: safeAreaTop
+                    });
+                    pressablesText = formatScreenStateSummary(screenSpaceSs, (p) => ({
+                        center: { x: toPx(p.center.x), y: toPx(p.center.y) },
+                        frame: {
+                            x: toPx(p.bounds.x),
+                            y: toPx(p.bounds.y),
+                            width: toPx(p.bounds.width),
+                            height: toPx(p.bounds.height),
+                        },
+                    }));
+                    pressablesIsScreenState = true;
                 }
             }
             catch {
             }
-            if (pointWidth === 0) {
-                pointWidth = Math.round(pixelWidth / scaleFactor);
-                pointHeight = Math.round(pixelHeight / scaleFactor);
+        }
+        const deliveredWidth = result.scaleFactor && result.scaleFactor > 1
+            ? Math.round(pixelWidth / result.scaleFactor)
+            : pixelWidth;
+        const deliveredHeight = result.scaleFactor && result.scaleFactor > 1
+            ? Math.round(pixelHeight / result.scaleFactor)
+            : pixelHeight;
+        let infoText;
+        if (result.scaleFactor && result.scaleFactor > 1) {
+            infoText = `Screenshot: raw ${pixelWidth}x${pixelHeight} px → delivered ${deliveredWidth}x${deliveredHeight} px (downscaled ${(1 / result.scaleFactor).toFixed(3)}× to fit API limits). Pressable coordinates below are in delivered-image pixels.`;
+        }
+        else {
+            infoText = `Screenshot captured (${pixelWidth}x${pixelHeight} pixels)`;
+        }
+        if (resolved.udid) {
+            recordScreenMetrics(resolved.udid, {
+                rawWidth: pixelWidth,
+                rawHeight: pixelHeight,
+                deliveredWidth,
+                deliveredHeight,
+                downscale: result.scaleFactor && result.scaleFactor > 1 ? 1 / result.scaleFactor : 1,
+                pointWidth,
+                pointHeight,
+                scale: scaleFactor,
+                capturedAt: Date.now(),
+            });
+        }
+        if (targetUdid) {
+            infoText += `\n📸 Captured from: ${targetDeviceName ? `${targetDeviceName} ` : ""}(${targetUdid})`;
+        }
+        infoText += `\n📱 iOS screen: ${pointWidth}x${pointHeight} points (${scaleFactor}x scale)`;
+        infoText += `\n📐 tap() handles pixel-to-point conversion automatically — pass pixel coords from this image directly`;
+        infoText += `\n⚠️ Status bar + safe area: ${safeAreaTop} points (${safeAreaOffsetPixels} pixels) from top`;
+        if (pressablesText) {
+            infoText += pressablesIsScreenState
+                ? `\n\n🧭 Screen state (route + tappable elements, coordinates in screenshot pixels):\n`
+                : `\n\n🎯 Pressable elements (ready-to-tap, coordinates in screenshot pixels):`;
+            infoText += `\n${pressablesText}`;
+            infoText += `\n\n💡 Next steps:`;
+            infoText += `\n  • tap(text="Button Label") — when text is exact and unique`;
+            infoText += `\n  • tap(testID="id") or tap(component="Name") — when you know the identifier`;
+            infoText += `\n  • tap(x=<px>, y=<px>) — use coordinates from the pressable elements list above (reliable for icons and ambiguous elements)`;
+            infoText += `\n  • get_screen_layout — full component tree when you need more than pressables`;
+        }
+        else {
+            if (!targetApp && connectedApps.size > 0) {
+                infoText += `\n\nℹ️ Pressable enrichment skipped: no RN app is connected to simulator ${resolvedUdid}.`;
+                infoText += ` ${connectedApps.size} other app(s) are connected on different device(s) — their fiber data was intentionally not used to avoid mismatched output.`;
             }
-            const safeAreaOffsetPixels = safeAreaTop * scaleFactor;
-            let pressablesText = null;
-            let pressablesIsScreenState = false;
-            if (targetApp) {
-                try {
-                    const ssResult = await screenStatePromise;
-                    if (ssResult && ssResult.success && ssResult.screenState) {
-                        const screenshotScale = result.scaleFactor || 1;
-                        const toPx = (v) => Math.round((v * scaleFactor) / screenshotScale);
-                        const screenSpaceSs = screenStateToScreenSpace(ssResult.screenState, {
-                            platform: "ios",
-                            topInset: safeAreaTop
-                        });
-                        pressablesText = formatScreenStateSummary(screenSpaceSs, (p) => ({
-                            center: { x: toPx(p.center.x), y: toPx(p.center.y) },
-                            frame: {
-                                x: toPx(p.bounds.x),
-                                y: toPx(p.bounds.y),
-                                width: toPx(p.bounds.width),
-                                height: toPx(p.bounds.height),
-                            },
-                        }));
-                        pressablesIsScreenState = true;
-                    }
-                }
-                catch {
+            infoText += `\n\n💡 Next steps:`;
+            infoText += `\n  • tap(text="Button Label") — tap element by visible text`;
+            infoText += `\n  • tap(x=<px>, y=<px>) — tap at coordinates from this screenshot`;
+            infoText += `\n  • get_screen_layout — get full UI tree with real on-screen positions`;
+        }
+        if (targetApp) {
+            try {
+                const logBoxState = await detectLogBox(targetDeviceName);
+                if (logBoxState && logBoxState.total > 0) {
+                    infoText += formatLogBoxWarning(logBoxState);
                 }
             }
-            const deliveredWidth = result.scaleFactor && result.scaleFactor > 1
-                ? Math.round(pixelWidth / result.scaleFactor)
-                : pixelWidth;
-            const deliveredHeight = result.scaleFactor && result.scaleFactor > 1
-                ? Math.round(pixelHeight / result.scaleFactor)
-                : pixelHeight;
-            let infoText;
-            if (result.scaleFactor && result.scaleFactor > 1) {
-                infoText = `Screenshot: raw ${pixelWidth}x${pixelHeight} px → delivered ${deliveredWidth}x${deliveredHeight} px (downscaled ${(1 / result.scaleFactor).toFixed(3)}× to fit API limits). Pressable coordinates below are in delivered-image pixels.`;
+            catch {
             }
-            else {
-                infoText = `Screenshot captured (${pixelWidth}x${pixelHeight} pixels)`;
+        }
+        try {
+            const overlay = await detectIOSSystemOverlay(resolvedUdid ?? undefined);
+            if (overlay) {
+                infoText += formatIOSSystemOverlayWarning(overlay);
             }
-            if (resolved.udid) {
-                recordScreenMetrics(resolved.udid, {
-                    rawWidth: pixelWidth,
-                    rawHeight: pixelHeight,
-                    deliveredWidth,
-                    deliveredHeight,
-                    downscale: result.scaleFactor && result.scaleFactor > 1 ? 1 / result.scaleFactor : 1,
-                    pointWidth,
-                    pointHeight,
-                    scale: scaleFactor,
-                    capturedAt: Date.now(),
-                });
+        }
+        catch {
+        }
+        imageBuffer.add({
+            id: `ios-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`,
+            image: result.data,
+            timestamp: Date.now(),
+            source: "ios_screenshot",
+            metadata: {
+                width: result.originalWidth || 0,
+                height: result.originalHeight || 0,
+                scaleFactor: result.scaleFactor || 1,
+                platform: "ios",
+            },
+        });
+        return {
+            content: [
+                {
+                    type: "text",
+                    text: infoText
+                },
+                {
+                    type: "image",
+                    data: result.data.toString("base64"),
+                    mimeType: "image/jpeg"
+                }
+            ]
+        };
+    }
+    return {
+        content: [
+            {
+                type: "text",
+                text: `Screenshot saved to: ${result.result}`
             }
-            if (targetUdid) {
-                infoText += `\n📸 Captured from: ${targetDeviceName ? `${targetDeviceName} ` : ""}(${targetUdid})`;
+        ]
+    };
+}
+async function androidScreenshotHandler({ outputPath, deviceId: deviceIdArg, device }) {
+    const deviceId = deviceIdArg ?? device;
+    const resolved = await resolveAndroidDeviceId(deviceId);
+    if (!resolved.ok)
+        return resolved.response;
+    const androidTargetApp = getConnectedAppByAndroidDeviceId(resolved.serial);
+    const capturePromise = androidScreenshot(outputPath, resolved.serial);
+    const statusBarPromise = androidGetStatusBarHeight(resolved.serial).catch(() => null);
+    const densityPromise = androidGetDensity(resolved.serial).catch(() => null);
+    const screenStatePromise = androidTargetApp
+        ? getScreenState({ device: androidTargetApp.deviceInfo.deviceName }).catch(() => null)
+        : Promise.resolve(null);
+    const result = await capturePromise;
+    if (!result.success) {
+        return {
+            content: [
+                {
+                    type: "text",
+                    text: `Error: ${result.error}`
+                }
+            ],
+            isError: true
+        };
+    }
+    if (result.data) {
+        const pixelWidth = result.originalWidth || 0;
+        const pixelHeight = result.originalHeight || 0;
+        const targetApp = getConnectedAppByAndroidDeviceId(resolved.serial);
+        const targetDeviceName = targetApp?.deviceInfo.deviceName;
+        if (targetApp) {
+            targetApp.lastScreenshot = {
+                originalWidth: pixelWidth,
+                originalHeight: pixelHeight,
+                scaleFactor: result.scaleFactor || 1,
+            };
+        }
+        const androidDeliveredW = result.scaleFactor && result.scaleFactor > 1
+            ? Math.round(pixelWidth / result.scaleFactor)
+            : pixelWidth;
+        const androidDeliveredH = result.scaleFactor && result.scaleFactor > 1
+            ? Math.round(pixelHeight / result.scaleFactor)
+            : pixelHeight;
+        let infoText = result.scaleFactor && result.scaleFactor > 1
+            ? `Screenshot: raw ${pixelWidth}x${pixelHeight} px → delivered ${androidDeliveredW}x${androidDeliveredH} px (downscaled ${(1 / result.scaleFactor).toFixed(3)}× to fit API limits). Pressable coordinates below are in delivered-image pixels.`
+            : `Screenshot captured (${pixelWidth}x${pixelHeight} pixels)`;
+        if (resolved.serial) {
+            recordScreenMetrics(resolved.serial, {
+                rawWidth: pixelWidth,
+                rawHeight: pixelHeight,
+                deliveredWidth: androidDeliveredW,
+                deliveredHeight: androidDeliveredH,
+                downscale: result.scaleFactor && result.scaleFactor > 1 ? 1 / result.scaleFactor : 1,
+                capturedAt: Date.now(),
+            });
+        }
+        let statusBarPixels = 63;
+        let statusBarDp = 24;
+        let densityDpi = 440;
+        try {
+            const statusBarResult = await statusBarPromise;
+            if (statusBarResult && statusBarResult.success && statusBarResult.heightPixels) {
+                statusBarPixels = statusBarResult.heightPixels;
+                statusBarDp = statusBarResult.heightDp || 24;
             }
-            infoText += `\n📱 iOS screen: ${pointWidth}x${pointHeight} points (${scaleFactor}x scale)`;
-            infoText += `\n📐 tap() handles pixel-to-point conversion automatically — pass pixel coords from this image directly`;
-            infoText += `\n⚠️ Status bar + safe area: ${safeAreaTop} points (${safeAreaOffsetPixels} pixels) from top`;
-            if (pressablesText) {
-                infoText += pressablesIsScreenState
-                    ? `\n\n🧭 Screen state (route + tappable elements, coordinates in screenshot pixels):\n`
-                    : `\n\n🎯 Pressable elements (ready-to-tap, coordinates in screenshot pixels):`;
-                infoText += `\n${pressablesText}`;
-                infoText += `\n\n💡 Next steps:`;
-                infoText += `\n  • tap(text="Button Label") — when text is exact and unique`;
-                infoText += `\n  • tap(testID="id") or tap(component="Name") — when you know the identifier`;
-                infoText += `\n  • tap(x=<px>, y=<px>) — use coordinates from the pressable elements list above (reliable for icons and ambiguous elements)`;
-                infoText += `\n  • get_screen_layout — full component tree when you need more than pressables`;
+            const densityResult = await densityPromise;
+            if (densityResult && densityResult.success && densityResult.density) {
+                densityDpi = densityResult.density;
             }
-            else {
-                if (!targetApp && connectedApps.size > 0) {
-                    infoText += `\n\nℹ️ Pressable enrichment skipped: no RN app is connected to simulator ${resolvedUdid}.`;
-                    infoText += ` ${connectedApps.size} other app(s) are connected on different device(s) — their fiber data was intentionally not used to avoid mismatched output.`;
+        }
+        catch {
+        }
+        let pressablesText = null;
+        let pressablesIsScreenState = false;
+        if (targetApp) {
+            try {
+                const ssResult = await screenStatePromise;
+                if (ssResult && ssResult.success && ssResult.screenState) {
+                    const screenshotScale = result.scaleFactor || 1;
+                    const densityScale = densityDpi / 160;
+                    const toPx = (v) => Math.round((v * densityScale) / screenshotScale);
+                    const screenSpaceSs = screenStateToScreenSpace(ssResult.screenState, {
+                        platform: "android",
+                        topInset: statusBarDp
+                    });
+                    pressablesText = formatScreenStateSummary(screenSpaceSs, (p) => ({
+                        center: { x: toPx(p.center.x), y: toPx(p.center.y) },
+                        frame: {
+                            x: toPx(p.bounds.x),
+                            y: toPx(p.bounds.y),
+                            width: toPx(p.bounds.width),
+                            height: toPx(p.bounds.height),
+                        },
+                    }));
+                    pressablesIsScreenState = true;
                 }
-                infoText += `\n\n💡 Next steps:`;
-                infoText += `\n  • tap(text="Button Label") — tap element by visible text`;
-                infoText += `\n  • tap(x=<px>, y=<px>) — tap at coordinates from this screenshot`;
-                infoText += `\n  • get_screen_layout — get full UI tree with real on-screen positions`;
             }
-            if (targetApp) {
-                try {
-                    const logBoxState = await detectLogBox(targetDeviceName);
-                    if (logBoxState && logBoxState.total > 0) {
-                        infoText += formatLogBoxWarning(logBoxState);
-                    }
-                }
-                catch {
-                }
+            catch {
             }
+        }
+        infoText += `\n📱 Android uses PIXELS for all coordinates`;
+        if (result.scaleFactor && result.scaleFactor > 1) {
+            infoText += `\n📐 tap() handles coordinate conversion automatically — pass pixel coords from this image directly`;
+        }
+        else {
+            infoText += `\n📐 Screenshot coords = tap coords (no conversion needed)`;
+        }
+        infoText += `\n⚠️ Status bar: ${statusBarPixels}px (${statusBarDp}dp) from top - app content starts below this`;
+        infoText += `\n📊 Display density: ${densityDpi}dpi`;
+        if (pressablesText) {
+            infoText += pressablesIsScreenState
+                ? `\n\n🧭 Screen state (route + tappable elements, coordinates in screenshot pixels):\n`
+                : `\n\n🎯 Pressable elements (ready-to-tap, coordinates in screenshot pixels):`;
+            infoText += `\n${pressablesText}`;
+            infoText += `\n\n💡 Next steps:`;
+            infoText += `\n  • tap(text="Button Label") — when text is exact and unique`;
+            infoText += `\n  • tap(testID="id") or tap(component="Name") — when you know the identifier`;
+            infoText += `\n  • tap(x=<px>, y=<px>) — use coordinates from the pressable elements list above (reliable for icons and ambiguous elements)`;
+            infoText += `\n  • get_screen_layout — full component tree when you need more than pressables`;
+        }
+        else {
+            if (!targetApp && connectedApps.size > 0) {
+                infoText += `\n\nℹ️ Pressable enrichment skipped: no RN app is connected to device ${deviceId ?? "(default)"}.`;
+                infoText += ` ${connectedApps.size} other app(s) are connected on different device(s) — their fiber data was intentionally not used to avoid mismatched output.`;
+            }
+            infoText += `\n\n💡 Next steps:`;
+            infoText += `\n  • tap(text="Button Label") — tap element by visible text`;
+            infoText += `\n  • tap(x=<px>, y=<px>) — tap at coordinates from this screenshot`;
+            infoText += `\n  • get_screen_layout — get full UI tree with real on-screen positions`;
+        }
+        if (targetApp) {
             try {
-                const overlay = await detectIOSSystemOverlay(resolvedUdid ?? undefined);
-                if (overlay) {
-                    infoText += formatIOSSystemOverlayWarning(overlay);
+                const logBoxState = await detectLogBox(targetDeviceName);
+                if (logBoxState && logBoxState.total > 0) {
+                    infoText += formatLogBoxWarning(logBoxState);
                 }
             }
             catch {
             }
-            imageBuffer.add({
-                id: `ios-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`,
-                image: result.data,
-                timestamp: Date.now(),
-                source: "ios_screenshot",
-                metadata: {
-                    width: result.originalWidth || 0,
-                    height: result.originalHeight || 0,
-                    scaleFactor: result.scaleFactor || 1,
-                    platform: "ios",
-                },
-            });
-            return {
-                content: [
-                    {
-                        type: "text",
-                        text: infoText
-                    },
-                    {
-                        type: "image",
-                        data: result.data.toString("base64"),
-                        mimeType: "image/jpeg"
-                    }
-                ]
-            };
         }
+        imageBuffer.add({
+            id: `android-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`,
+            image: result.data,
+            timestamp: Date.now(),
+            source: "android_screenshot",
+            metadata: {
+                width: result.originalWidth || 0,
+                height: result.originalHeight || 0,
+                scaleFactor: result.scaleFactor || 1,
+                platform: "android",
+            },
+        });
         return {
             content: [
                 {
                     type: "text",
-                    text: `Screenshot saved to: ${result.result}`
+                    text: infoText
+                },
+                {
+                    type: "image",
+                    data: result.data.toString("base64"),
+                    mimeType: "image/jpeg"
                 }
             ]
         };
+    }
+    return {
+        content: [
+            {
+                type: "text",
+                text: `Screenshot saved to: ${result.result}`
+            }
+        ]
+    };
+}
+async function chromiumScreenshotResponse(app, outputPath) {
+    let shot;
+    try {
+        shot = await chromiumCapture(app);
+    }
+    catch (err) {
+        return {
+            content: [{ type: "text", text: `Error: ${err instanceof Error ? err.message : String(err)}` }],
+            isError: true,
+        };
+    }
+    let saveNote = "";
+    if (outputPath) {
+        try {
+            writeFileSync(outputPath, shot.buffer);
+        }
+        catch (err) {
+            saveNote = `\nCould not save to ${outputPath}: ${err instanceof Error ? err.message : String(err)}`;
+        }
+    }
+    imageBuffer.add({
+        id: `chromium-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`,
+        image: shot.buffer,
+        timestamp: Date.now(),
+        source: "screenshot",
+        metadata: { width: shot.width, height: shot.height, scaleFactor: shot.scaleFactor, platform: "chromium" },
     });
+    const summary = await chromiumScreenState(app).then((ss) => formatScreenStateSummary(ss)).catch(() => null);
+    const { w, h, dpr } = shot.viewport;
+    const downscaled = shot.scaleFactor !== 1
+        ? `, downscaled from ${Math.round(w * dpr)}x${Math.round(h * dpr)} to fit API limits`
+        : "";
+    const text = `Screenshot ${shot.width}x${shot.height} px of ${app.deviceInfo.deviceName} ` +
+        `(chromium viewport ${w}x${h} CSS px at devicePixelRatio ${dpr}${downscaled}).\n` +
+        `Pass pixel coordinates from this image straight to tap(x, y, device).` + saveNote +
+        (summary ? `\n\n${summary}` : "");
+    return {
+        content: [
+            { type: "text", text },
+            { type: "image", data: shot.buffer.toString("base64"), mimeType: "image/jpeg" },
+        ],
+    };
+}
+export function registerScreenshotTools(server) {
+    registerToolWithTelemetry(server, "ios_screenshot", {
+        description: "Take a screenshot from an iOS simulator. Returns the image plus a screen-state summary: active route (name + navigation stack), overlay-grouped tappable elements (pressables behind an open sheet/modal are excluded), component names as JSX tags, labels, testIDs, and frames — all in ready-to-tap pixel coordinates. Prefer tap(text=\"...\") when text is exact and unique; otherwise use tap(x, y) with coordinates from the list — this is the most reliable way to tap icons or visually-identified elements. Use component names for inspect_component/find_components.\n" +
+            "PURPOSE: Snapshot what the user sees on iOS AND get tap-ready pressables + a structured component map in one call.\n" +
+            "WHEN TO USE: Any visual verification, before/after comparison, or as the starting point for tapping UI by coordinates.\n" +
+            "WORKFLOW: ios_screenshot -> pick element from pressables -> tap(x, y) or tap(testID=...) -> ios_screenshot to verify.\n" +
+            "PHYSICAL DEVICES: a USB-attached iPhone/iPad works too (pass its UDID or name from list_devices) — capture only, via pymobiledevice3. No pressables, no tap/swipe/input_text: iOS has no touch injection below 17.\n" +
+            "LIMITATIONS: Otherwise requires a booted iOS simulator (simctl). For system dialogs without RN, combine with tap(..., native=true).\n" +
+            "GOOD: ios_screenshot()\n" +
+            "BAD: ios_screenshot({ udid: \"guess\" }) with a made-up UDID — run list_devices first.\n" +
+            "SOURCE: to jump from a pixel to the code that renders it, call inspect_at_point(x, y) — it returns the absolute file and line.\n",
+        inputSchema: {
+            outputPath: z
+                .string()
+                .optional()
+                .describe("Optional path to save the screenshot. If not provided, saves to temp directory."),
+            udid: z
+                .string()
+                .optional()
+                .describe(IOS_ARG_DESC),
+            device: z
+                .string()
+                .optional()
+                .describe("Alias for `udid` — same accepted values. Provided for consistency with tap/get_screen_layout/get_screen_state, which all use `device`. If both are given, `udid` wins.")
+        }
+    }, iosScreenshotHandler);
     registerToolWithTelemetry(server, "android_screenshot", {
         description: "Take a screenshot from an Android device/emulator. Returns the image plus a screen-state summary: active route (name + navigation stack), overlay-grouped tappable elements (pressables behind an open sheet/modal are excluded), component names as JSX tags, labels, testIDs, and frames — all in ready-to-tap pixel coordinates. Prefer tap(text=\"...\") when text is exact and unique; otherwise use tap(x, y) with coordinates from the list — this is the most reliable way to tap icons or visually-identified elements. Use component names for inspect_component/find_components.\n" +
             "PURPOSE: Snapshot what the user sees on Android AND get tap-ready pressables + a structured component map in one call.\n" +
@@ -303,179 +526,35 @@ export function registerScreenshotTools(server) {
                 .optional()
                 .describe("Alias for `deviceId` — same accepted values. Provided for consistency with tap/get_screen_state/ios_screenshot, which all accept `device`. If both are given, `deviceId` wins.")
         }
-    }, async ({ outputPath, deviceId: deviceIdArg, device }) => {
-        const deviceId = deviceIdArg ?? device;
-        const resolved = await resolveAndroidDeviceId(deviceId);
-        if (!resolved.ok)
-            return resolved.response;
-        const androidTargetApp = getConnectedAppByAndroidDeviceId(resolved.serial);
-        const capturePromise = androidScreenshot(outputPath, resolved.serial);
-        const statusBarPromise = androidGetStatusBarHeight(resolved.serial).catch(() => null);
-        const densityPromise = androidGetDensity(resolved.serial).catch(() => null);
-        const screenStatePromise = androidTargetApp
-            ? getScreenState({ device: androidTargetApp.deviceInfo.deviceName }).catch(() => null)
-            : Promise.resolve(null);
-        const result = await capturePromise;
-        if (!result.success) {
-            return {
-                content: [
-                    {
-                        type: "text",
-                        text: `Error: ${result.error}`
-                    }
-                ],
-                isError: true
-            };
-        }
-        if (result.data) {
-            const pixelWidth = result.originalWidth || 0;
-            const pixelHeight = result.originalHeight || 0;
-            const targetApp = getConnectedAppByAndroidDeviceId(resolved.serial);
-            const targetDeviceName = targetApp?.deviceInfo.deviceName;
-            if (targetApp) {
-                targetApp.lastScreenshot = {
-                    originalWidth: pixelWidth,
-                    originalHeight: pixelHeight,
-                    scaleFactor: result.scaleFactor || 1,
-                };
-            }
-            const androidDeliveredW = result.scaleFactor && result.scaleFactor > 1
-                ? Math.round(pixelWidth / result.scaleFactor)
-                : pixelWidth;
-            const androidDeliveredH = result.scaleFactor && result.scaleFactor > 1
-                ? Math.round(pixelHeight / result.scaleFactor)
-                : pixelHeight;
-            let infoText = result.scaleFactor && result.scaleFactor > 1
-                ? `Screenshot: raw ${pixelWidth}x${pixelHeight} px → delivered ${androidDeliveredW}x${androidDeliveredH} px (downscaled ${(1 / result.scaleFactor).toFixed(3)}× to fit API limits). Pressable coordinates below are in delivered-image pixels.`
-                : `Screenshot captured (${pixelWidth}x${pixelHeight} pixels)`;
-            if (resolved.serial) {
-                recordScreenMetrics(resolved.serial, {
-                    rawWidth: pixelWidth,
-                    rawHeight: pixelHeight,
-                    deliveredWidth: androidDeliveredW,
-                    deliveredHeight: androidDeliveredH,
-                    downscale: result.scaleFactor && result.scaleFactor > 1 ? 1 / result.scaleFactor : 1,
-                    capturedAt: Date.now(),
-                });
-            }
-            let statusBarPixels = 63;
-            let statusBarDp = 24;
-            let densityDpi = 440;
-            try {
-                const statusBarResult = await statusBarPromise;
-                if (statusBarResult && statusBarResult.success && statusBarResult.heightPixels) {
-                    statusBarPixels = statusBarResult.heightPixels;
-                    statusBarDp = statusBarResult.heightDp || 24;
-                }
-                const densityResult = await densityPromise;
-                if (densityResult && densityResult.success && densityResult.density) {
-                    densityDpi = densityResult.density;
-                }
-            }
-            catch {
-            }
-            let pressablesText = null;
-            let pressablesIsScreenState = false;
-            if (targetApp) {
-                try {
-                    const ssResult = await screenStatePromise;
-                    if (ssResult && ssResult.success && ssResult.screenState) {
-                        const screenshotScale = result.scaleFactor || 1;
-                        const densityScale = densityDpi / 160;
-                        const toPx = (v) => Math.round((v * densityScale) / screenshotScale);
-                        const screenSpaceSs = screenStateToScreenSpace(ssResult.screenState, {
-                            platform: "android",
-                            topInset: statusBarDp
-                        });
-                        pressablesText = formatScreenStateSummary(screenSpaceSs, (p) => ({
-                            center: { x: toPx(p.center.x), y: toPx(p.center.y) },
-                            frame: {
-                                x: toPx(p.bounds.x),
-                                y: toPx(p.bounds.y),
-                                width: toPx(p.bounds.width),
-                                height: toPx(p.bounds.height),
-                            },
-                        }));
-                        pressablesIsScreenState = true;
-                    }
-                }
-                catch {
-                }
-            }
-            infoText += `\n📱 Android uses PIXELS for all coordinates`;
-            if (result.scaleFactor && result.scaleFactor > 1) {
-                infoText += `\n📐 tap() handles coordinate conversion automatically — pass pixel coords from this image directly`;
-            }
-            else {
-                infoText += `\n📐 Screenshot coords = tap coords (no conversion needed)`;
-            }
-            infoText += `\n⚠️ Status bar: ${statusBarPixels}px (${statusBarDp}dp) from top - app content starts below this`;
-            infoText += `\n📊 Display density: ${densityDpi}dpi`;
-            if (pressablesText) {
-                infoText += pressablesIsScreenState
-                    ? `\n\n🧭 Screen state (route + tappable elements, coordinates in screenshot pixels):\n`
-                    : `\n\n🎯 Pressable elements (ready-to-tap, coordinates in screenshot pixels):`;
-                infoText += `\n${pressablesText}`;
-                infoText += `\n\n💡 Next steps:`;
-                infoText += `\n  • tap(text="Button Label") — when text is exact and unique`;
-                infoText += `\n  • tap(testID="id") or tap(component="Name") — when you know the identifier`;
-                infoText += `\n  • tap(x=<px>, y=<px>) — use coordinates from the pressable elements list above (reliable for icons and ambiguous elements)`;
-                infoText += `\n  • get_screen_layout — full component tree when you need more than pressables`;
-            }
-            else {
-                if (!targetApp && connectedApps.size > 0) {
-                    infoText += `\n\nℹ️ Pressable enrichment skipped: no RN app is connected to device ${deviceId ?? "(default)"}.`;
-                    infoText += ` ${connectedApps.size} other app(s) are connected on different device(s) — their fiber data was intentionally not used to avoid mismatched output.`;
-                }
-                infoText += `\n\n💡 Next steps:`;
-                infoText += `\n  • tap(text="Button Label") — tap element by visible text`;
-                infoText += `\n  • tap(x=<px>, y=<px>) — tap at coordinates from this screenshot`;
-                infoText += `\n  • get_screen_layout — get full UI tree with real on-screen positions`;
-            }
-            if (targetApp) {
-                try {
-                    const logBoxState = await detectLogBox(targetDeviceName);
-                    if (logBoxState && logBoxState.total > 0) {
-                        infoText += formatLogBoxWarning(logBoxState);
-                    }
-                }
-                catch {
-                }
-            }
-            imageBuffer.add({
-                id: `android-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`,
-                image: result.data,
-                timestamp: Date.now(),
-                source: "android_screenshot",
-                metadata: {
-                    width: result.originalWidth || 0,
-                    height: result.originalHeight || 0,
-                    scaleFactor: result.scaleFactor || 1,
-                    platform: "android",
-                },
-            });
+    }, androidScreenshotHandler);
+    registerToolWithTelemetry(server, "screenshot", {
+        description: "Take a screenshot of whichever target `device` resolves to: an iOS simulator, an Android device, or a chromium (Electron / Chrome) window.\n" +
+            "PURPOSE: One capture tool for every platform. On iOS and Android it is exactly ios_screenshot / android_screenshot, pressables summary included. On chromium it captures the page viewport (no window chrome) over CDP, with the same element summary as get_screen_state.\n" +
+            "COORDINATES: pixels in the returned image are the coordinates tap(x, y) takes, on every platform. Never scale them yourself.\n" +
+            "GOOD: screenshot(); screenshot({ device: \"FluentTalk\" })\n" +
+            "LIMITATIONS: a hidden or minimised chromium window does not paint and cannot be captured: the call fails at once and says so.",
+        inputSchema: {
+            device: z
+                .string()
+                .optional()
+                .describe("Target: a connected app name (get_apps), or a simulator/emulator name, UDID or adb serial. Omit for the default target."),
+            outputPath: z.string().optional().describe("Optional path to also save the image to."),
+        },
+    }, async ({ device, outputPath }) => {
+        const chromeApp = chromiumAppFor("screenshot", device);
+        if (chromeApp)
+            return await chromiumScreenshotResponse(chromeApp, outputPath);
+        const resolved = await resolveDeviceTarget(device);
+        if (!resolved.ok) {
             return {
-                content: [
-                    {
-                        type: "text",
-                        text: infoText
-                    },
-                    {
-                        type: "image",
-                        data: result.data.toString("base64"),
-                        mimeType: "image/jpeg"
-                    }
-                ]
+                content: [{ type: "text", text: `Error: ${formatResolverError(resolved.error)}` }],
+                isError: true,
             };
         }
-        return {
-            content: [
-                {
-                    type: "text",
-                    text: `Screenshot saved to: ${result.result}`
-                }
-            ]
-        };
+        const t = resolved.target;
+        return t.platform === "ios"
+            ? await iosScreenshotHandler({ outputPath, udid: t.iosUdid ?? device })
+            : await androidScreenshotHandler({ outputPath, deviceId: t.androidSerial ?? device });
     });
     registerToolWithTelemetry(server, "get_images", {
         description: "Access the shared image buffer containing screenshots from all tools (ios_screenshot, android_screenshot, tap verification). Returns metadata only by default — use id or groupId+frameIndex to retrieve actual image data. Tap burst verification stores frame groups here when burst=true is used.\n" +