safari-mcp 2.17.1 → 2.18.0

package/extension/background.js

@@ -61,6 +61,30 @@ async function _bridgeFetch(url, init = {}) {
   headers.set("X-Safari-MCP-Worker", _bridgeWorkerId);
   return fetch(url, { ...init, headers });
 }
+const _CONTENT_WAKE_TOKEN_STORAGE_KEY = "mcpContentWakeTokenV1";
+const _CONTENT_WAKE_LABEL = "safari-mcp-content-wakeup-v1";
+async function _publishContentWakeToken() {
+  try {
+    const bridgeToken = await _bridgeAuthToken();
+    const encoder = new TextEncoder();
+    const key = await crypto.subtle.importKey(
+      "raw",
+      encoder.encode(bridgeToken),
+      { name: "HMAC", hash: "SHA-256" },
+      false,
+      ["sign"]
+    );
+    const signature = await crypto.subtle.sign("HMAC", key, encoder.encode(_CONTENT_WAKE_LABEL));
+    const wakeToken = Array.from(new Uint8Array(signature), (byte) =>
+      byte.toString(16).padStart(2, "0")
+    ).join("");
+    if (!/^[0-9a-f]{64}$/.test(wakeToken)) throw new Error("invalid content wake token");
+    await browser.storage.local.set({ [_CONTENT_WAKE_TOKEN_STORAGE_KEY]: wakeToken });
+  } catch (error) {
+    await browser.storage.local.remove(_CONTENT_WAKE_TOKEN_STORAGE_KEY).catch(() => {});
+    throw error;
+  }
+}
 let _targetProfile = null;   // Profile name from server (e.g. "Automations")
 let _profileWindowId = null; // Discovered windowId for the profile (shared default)
 // Per-session profile window. One Safari profile can hold SEVERAL windows once
@@ -419,6 +443,18 @@ let _startupReady = browser.storage.local.get("mcpEnabled").then(data => {
 
 // Listen for messages from popup
 browser.runtime.onMessage.addListener((msg, sender, sendResponse) => {
+  // Safari can suspend an MV3 worker even while a content-script Port exists. A
+  // leader page therefore sends this fixed, authority-free message as a second
+  // keepalive signal. runtime.sendMessage is an extension event, so it wakes the
+  // worker on Safari versions where Port traffic alone does not. Never accept tab,
+  // URL, receipt, session, or command data through this path.
+  if (msg?.action === "mcpContentKeepalivePingV1") {
+    if (Object.keys(msg).length !== 1) return false;
+    if (!Number.isInteger(sender?.tab?.id)) return false;
+    if (!_enabled || _bridgeWorkerSuperseded || _bridgeWorkerRetiring) return false;
+    sendResponse({ ok: true });
+    return false;
+  }
   if (msg.action === "setEnabled") {
     _enabled = msg.enabled;
     if (!_enabled) {
@@ -545,6 +581,14 @@ async function connect() {
         signal: AbortSignal.timeout(1500),
       });
       if (_restartInvalidatedConnect(pollGeneration)) return;
+      if (res.status === 423) {
+        isConnected = false;
+        updateBadge("");
+        _stopHeartbeat();
+        scheduleReconnect(3000);
+        _connecting = false;
+        return;
+      }
       if (!res.ok) continue;
 
       const data = await res.json().catch(() => ({}));
@@ -569,7 +613,7 @@ async function connect() {
             isConnected = false;
             updateBadge("");
             _stopHeartbeat();
-            scheduleReconnect();
+            scheduleReconnect(3000);
             _connecting = false;
             return;
           }
@@ -604,17 +648,22 @@ async function connect() {
   _connecting = false;
 }
 
-function scheduleReconnect() {
+function scheduleReconnect(delayOverrideMs = null) {
   if (!_enabled || _bridgeWorkerSuperseded || _bridgeWorkerRetiring) return;
   // Cancel any existing reconnect to prevent exponential growth
   if (_reconnectTimer) { clearTimeout(_reconnectTimer); _reconnectTimer = null; }
 
   // Single timer with exponential backoff (3s → 6s → 12s → ... → 60s max)
+  const delay = Number.isFinite(delayOverrideMs)
+    ? Math.max(0, Number(delayOverrideMs))
+    : _reconnectDelay;
   _reconnectTimer = setTimeout(() => {
     _reconnectTimer = null;
     connect();
-  }, _reconnectDelay);
-  _reconnectDelay = Math.min(_reconnectDelay * 2, _RECONNECT_MAX);
+  }, delay);
+  if (!Number.isFinite(delayOverrideMs)) {
+    _reconnectDelay = Math.min(_reconnectDelay * 2, _RECONNECT_MAX);
+  }
 
   // Alarm as backup — wakes terminated service worker (Safari minimum 1 minute)
   try {
@@ -1038,11 +1087,17 @@ async function handleCommand(type, payload) {
         );
         if (!r || r.ok !== true) throw new Error((r && r.error) || "content bridge failed");
         _injectionBlockedTabs.add(evalTabId);
+        if (evalOrigin) _injectionBlockedOrigins.set(evalOrigin, Date.now() + _INJECTION_BLOCK_TTL_MS);
         return r.value;
       };
-      if (_injectionBlockedTabs.has(evalTabId)) {
+      const evalOrigin = _originKey(targetTab?.url);
+      const originBlocked = !!evalOrigin && (_injectionBlockedOrigins.get(evalOrigin) || 0) > Date.now();
+      if (_injectionBlockedTabs.has(evalTabId) || originBlocked) {
         try { return await viaBridge(); }
-        catch (_e) { _injectionBlockedTabs.delete(evalTabId); } // page changed — re-probe
+        catch (_e) { // page changed — re-probe
+          _injectionBlockedTabs.delete(evalTabId);
+          if (evalOrigin) _injectionBlockedOrigins.delete(evalOrigin);
+        }
       }
 
       // First contact with a hardened page: every execInTab strategy below will stall.
@@ -1153,14 +1208,23 @@ async function handleCommand(type, payload) {
       // If it fails, fall back to AppleScript screencapture -l (which also doesn't steal focus).
       // NEVER use browser.windows.update({ focused: true }) — it steals user's keyboard/mouse.
       let captureWindowId = _profileWindowId || null;
+      // captureVisibleTab only sees the window's selected tab, so ours must be selected for
+      // the capture — but the selection belongs to the user when they browse this window.
+      // Remember what they had selected and put it back (the AppleScript path already does
+      // this via _withTargetTabFronted; this path used to leave our tab in front).
+      let restoreTabId = null;
       if (tabId) {
         try {
           const tabInfo = await browser.tabs.get(tabId);
           captureWindowId = tabInfo.windowId;
-          // Only activate the correct tab — does NOT bring Safari window to foreground
-          await browser.tabs.update(tabId, { active: true });
+          if (!tabInfo.active) {
+            const [selected] = await browser.tabs.query({ active: true, windowId: tabInfo.windowId });
+            if (selected && selected.id !== tabId) restoreTabId = selected.id;
+            // Only selects the tab — does NOT bring the Safari window to the foreground
+            await browser.tabs.update(tabId, { active: true });
+            await new Promise(r => setTimeout(r, 150));
+          }
         } catch (_) {}
-        await new Promise(r => setTimeout(r, 150));
       }
       // Use JPEG with quality 50 to reduce size (~600KB PNG → ~60KB JPEG)
       try {
@@ -1178,6 +1242,8 @@ async function handleCommand(type, payload) {
         }
         // Any other error also falls back — better than stealing focus
         return "__SCREENSHOT_PERMISSION_DENIED__";
+      } finally {
+        if (restoreTabId !== null) browser.tabs.update(restoreTabId, { active: true }).catch(() => {});
       }
     }
 
@@ -3668,6 +3734,15 @@ const MAIN_WORLD_INJECT_MS = 3000;
 // pages fast instead of merely working. Cleared when the tab navigates or closes, so a
 // page that stops blocking injection is re-probed rather than downgraded forever.
 const _injectionBlockedTabs = new Set();
+// Same fact, keyed by origin: the per-tab set is cleared on every navigation, so a hardened
+// SPA (business.facebook.com) re-paid the ~10s probe on the first evaluate after EACH
+// navigate — 356 evaluates ≥10s in one week. An origin that blocked injection once is sent
+// straight to the bridge for a few hours; a bridge failure clears it and re-probes.
+const _injectionBlockedOrigins = new Map(); // origin → expiry (ms)
+const _INJECTION_BLOCK_TTL_MS = 6 * 60 * 60 * 1000;
+function _originKey(url) {
+  try { return new URL(String(url || "")).origin; } catch { return ""; }
+}
 function _withInjectionDeadline(promise, ms = MAIN_WORLD_INJECT_MS) {
   let timer;
   return Promise.race([
@@ -4308,6 +4383,7 @@ console.log("Safari MCP Bridge: service worker started");
 updateBadge("");
 // Wait for storage to load before connecting (prevents race condition with _enabled)
 _startupReady.then(() => {
+  _publishContentWakeToken().catch(() => {});
   if (_enabled) connect();
   else updateBadge("OFF");
 }).catch(() => connect());

package/extension/command-content.js

@@ -32,6 +32,10 @@ const _MCP_KEEPALIVE_LEASE_MS = 26000;
 const _MCP_KEEPALIVE_RENEW_MS = 18000;
 const _MCP_KEEPALIVE_PING_MS = 8000;
 const _MCP_KEEPALIVE_RECONNECT_MS = 1500;
+const _MCP_CONTENT_WAKE_TOKEN_KEY = "mcpContentWakeTokenV1";
+const _MCP_CONTENT_WAKE_BRIDGE_KEY = "mcpBridgeUrl";
+const _MCP_CONTENT_WAKE_RETRY_MS = 1500;
+const _MCP_CONTENT_WAKE_PORTS = new Set(["9224", "9228", "9232", "9236"]);
 const _mcpKeepaliveInstanceId = inheritedKeepaliveId || (() => {
   const bytes = new Uint8Array(12);
   crypto.getRandomValues(bytes);
@@ -47,6 +51,11 @@ let _mcpKeepaliveEnabledRevision = 0;
 let _mcpKeepaliveLeader = false;
 let _mcpKeepaliveClaiming = false;
 let _mcpKeepaliveDisposed = false;
+let _mcpContentWakeToken = "";
+let _mcpContentWakeBridgeUrl = "";
+let _mcpContentWakeAbort = null;
+let _mcpContentWakeRetryTimer = null;
+let _mcpContentWakeGeneration = 0;
 
 function _isMcpKeepaliveTopFrame() {
   try { return window.top === window; } catch (_) { return false; }
@@ -134,9 +143,110 @@ function _clearMcpKeepalivePingTimer() {
   _mcpKeepalivePingTimer = null;
 }
 
+function _validMcpContentWakeToken(value) {
+  return /^[0-9a-f]{64}$/.test(String(value || "")) ? String(value) : "";
+}
+
+function _validMcpContentWakeBridgeUrl(value) {
+  try {
+    const url = new URL(String(value || ""));
+    if (url.protocol !== "http:") return "";
+    if (url.hostname !== "127.0.0.1" && url.hostname !== "localhost") return "";
+    if (!_MCP_CONTENT_WAKE_PORTS.has(url.port)) return "";
+    if (url.pathname !== "/" || url.search || url.hash) return "";
+    return url.origin;
+  } catch (_) {
+    return "";
+  }
+}
+
+function _sendMcpRuntimeWake() {
+  try {
+    const wakeup = browser.runtime.sendMessage({ action: "mcpContentKeepalivePingV1" });
+    wakeup?.catch?.(() => {});
+  } catch (_) {}
+}
+
+function _stopMcpContentWakeLoop() {
+  _mcpContentWakeGeneration += 1;
+  if (_mcpContentWakeRetryTimer) clearTimeout(_mcpContentWakeRetryTimer);
+  _mcpContentWakeRetryTimer = null;
+  const controller = _mcpContentWakeAbort;
+  _mcpContentWakeAbort = null;
+  if (controller) {
+    try { controller.abort(); } catch (_) {}
+  }
+}
+
+function _startMcpContentWakeLoop() {
+  if (
+    _mcpKeepaliveDisposed ||
+    !_mcpKeepaliveLeader ||
+    !_mcpKeepaliveEnabled ||
+    !_mcpContentWakeToken ||
+    !_mcpContentWakeBridgeUrl ||
+    _mcpContentWakeAbort ||
+    _mcpContentWakeRetryTimer
+  )
+    return;
+  const generation = ++_mcpContentWakeGeneration;
+  void _runMcpContentWakePoll(generation);
+}
+
+async function _runMcpContentWakePoll(generation) {
+  if (
+    generation !== _mcpContentWakeGeneration ||
+    _mcpKeepaliveDisposed ||
+    !_mcpKeepaliveLeader ||
+    !_mcpKeepaliveEnabled ||
+    !_mcpContentWakeToken ||
+    !_mcpContentWakeBridgeUrl
+  )
+    return;
+  const controller = new AbortController();
+  _mcpContentWakeAbort = controller;
+  let succeeded = false;
+  try {
+    const response = await fetch(`${_mcpContentWakeBridgeUrl}/content-wakeup`, {
+      cache: "no-store",
+      credentials: "omit",
+      headers: { "X-Safari-MCP-Wakeup": _mcpContentWakeToken },
+      mode: "cors",
+      referrerPolicy: "no-referrer",
+      signal: controller.signal,
+    });
+    succeeded = response.status === 204;
+  } catch (_) {
+    succeeded = false;
+  } finally {
+    if (_mcpContentWakeAbort === controller) _mcpContentWakeAbort = null;
+  }
+  if (
+    generation !== _mcpContentWakeGeneration ||
+    _mcpKeepaliveDisposed ||
+    !_mcpKeepaliveLeader ||
+    !_mcpKeepaliveEnabled
+  )
+    return;
+  if (succeeded) {
+    _sendMcpRuntimeWake();
+    void _runMcpContentWakePoll(generation);
+    return;
+  }
+  _mcpContentWakeRetryTimer = setTimeout(() => {
+    _mcpContentWakeRetryTimer = null;
+    if (generation !== _mcpContentWakeGeneration) return;
+    void _runMcpContentWakePoll(generation);
+  }, _MCP_CONTENT_WAKE_RETRY_MS);
+}
+
 function _sendMcpKeepalivePing() {
   if (_mcpKeepaliveDisposed || !_mcpKeepaliveLeader || !_mcpKeepaliveEnabled) return;
   _clearMcpKeepalivePingTimer();
+  // A direct runtime message wakes Safari's MV3 worker. The Port remains useful as
+  // a low-overhead continuous lease, but real Safari can suspend the worker despite
+  // an open Port; using both paths avoids the observed 30-second poll gaps.
+  _sendMcpRuntimeWake();
   const port = _openMcpKeepalivePort();
   if (port) {
     try { port.postMessage({ type: "ping" }); }
@@ -155,6 +265,7 @@ function _stopMcpKeepaliveLeadership() {
   if (_mcpKeepaliveReconnectTimer) clearTimeout(_mcpKeepaliveReconnectTimer);
   _mcpKeepaliveReconnectTimer = null;
   _clearMcpKeepalivePingTimer();
+  _stopMcpContentWakeLoop();
   _closeMcpKeepalivePort();
 }
 
@@ -183,6 +294,7 @@ function _becomeMcpKeepaliveLeader() {
   if (_mcpKeepaliveDisposed || !_mcpKeepaliveEnabled || !_isMcpKeepaliveTopFrame()) return;
   _mcpKeepaliveLeader = true;
   _clearMcpKeepaliveLeaseTimer();
+  _startMcpContentWakeLoop();
   _renewMcpKeepaliveLeaseAndPing();
 }
 
@@ -270,9 +382,20 @@ async function _initializeMcpKeepalive() {
   document.addEventListener("visibilitychange", _onMcpKeepaliveVisibilityChange);
   browser.storage.onChanged.addListener(_onMcpKeepaliveStorageChanged);
   const enabledRevision = _mcpKeepaliveEnabledRevision;
-  const stored = await browser.storage.local.get(["mcpEnabled", _MCP_KEEPALIVE_LEASE_KEY]).catch(() => ({}));
+  const stored = await browser.storage.local
+    .get([
+      "mcpEnabled",
+      _MCP_KEEPALIVE_LEASE_KEY,
+      _MCP_CONTENT_WAKE_TOKEN_KEY,
+      _MCP_CONTENT_WAKE_BRIDGE_KEY,
+    ])
+    .catch(() => ({}));
   if (_mcpKeepaliveDisposed || enabledRevision !== _mcpKeepaliveEnabledRevision) return;
   _mcpKeepaliveEnabled = stored.mcpEnabled !== false;
+  _mcpContentWakeToken = _validMcpContentWakeToken(stored[_MCP_CONTENT_WAKE_TOKEN_KEY]);
+  _mcpContentWakeBridgeUrl = _validMcpContentWakeBridgeUrl(
+    stored[_MCP_CONTENT_WAKE_BRIDGE_KEY]
+  );
   if (_mcpKeepaliveEnabled) {
     _considerMcpKeepaliveLeadership(stored[_MCP_KEEPALIVE_LEASE_KEY]);
   }
@@ -280,6 +403,20 @@ async function _initializeMcpKeepalive() {
 
 async function _onMcpKeepaliveStorageChanged(changes, areaName) {
   if (_mcpKeepaliveDisposed || areaName !== "local") return;
+  if (changes[_MCP_CONTENT_WAKE_TOKEN_KEY] || changes[_MCP_CONTENT_WAKE_BRIDGE_KEY]) {
+    _stopMcpContentWakeLoop();
+    if (changes[_MCP_CONTENT_WAKE_TOKEN_KEY]) {
+      _mcpContentWakeToken = _validMcpContentWakeToken(
+        changes[_MCP_CONTENT_WAKE_TOKEN_KEY].newValue
+      );
+    }
+    if (changes[_MCP_CONTENT_WAKE_BRIDGE_KEY]) {
+      _mcpContentWakeBridgeUrl = _validMcpContentWakeBridgeUrl(
+        changes[_MCP_CONTENT_WAKE_BRIDGE_KEY].newValue
+      );
+    }
+    if (_mcpKeepaliveLeader) _startMcpContentWakeLoop();
+  }
   if (changes.mcpEnabled) {
     _mcpKeepaliveEnabledRevision += 1;
     _mcpKeepaliveEnabled = changes.mcpEnabled.newValue !== false;

package/extension/manifest.json

@@ -2,25 +2,43 @@
   "manifest_version": 3,
   "name": "Safari MCP Bridge",
   "description": "Connects AI agents to Safari — fast JS execution, screenshots, clicks with your real cookies/logins",
-  "version": "2.10.4",
+  "version": "2.10.8",
   "icons": {
     "48": "images/icon-48.png",
     "96": "images/icon-96.png",
     "128": "images/icon-128.png"
   },
-  "background": { "service_worker": "background.js" },
+  "background": {
+    "scripts": [
+      "background.js"
+    ],
+    "persistent": false
+  },
   "action": {
     "default_popup": "popup/popup.html",
-    "default_icon": { "16": "images/icon-48.png", "32": "images/icon-48.png" }
+    "default_icon": {
+      "16": "images/icon-48.png",
+      "32": "images/icon-48.png"
+    }
   },
   "content_security_policy": {
     "extension_pages": "script-src 'self'; connect-src 'self' http://localhost:9224 http://127.0.0.1:9224 http://localhost:9228 http://127.0.0.1:9228 http://localhost:9232 http://127.0.0.1:9232 http://localhost:9236 http://127.0.0.1:9236"
   },
-  "permissions": ["activeTab", "scripting", "tabs", "alarms", "storage"],
-  "host_permissions": ["<all_urls>"],
+  "permissions": [
+    "activeTab",
+    "scripting",
+    "tabs",
+    "alarms",
+    "storage"
+  ],
+  "host_permissions": [
+    "<all_urls>"
+  ],
   "content_scripts": [
     {
-      "matches": ["<all_urls>"],
+      "matches": [
+        "<all_urls>"
+      ],
       "exclude_matches": [
         "*://*.linkedin.com/*",
         "*://*.x.com/*",
@@ -31,13 +49,19 @@
         "*://chat.google.com/*",
         "*://meet.google.com/*"
       ],
-      "js": ["content.js"],
+      "js": [
+        "content.js"
+      ],
       "run_at": "document_start",
       "world": "MAIN"
     },
     {
-      "matches": ["<all_urls>"],
-      "js": ["command-content.js"],
+      "matches": [
+        "<all_urls>"
+      ],
+      "js": [
+        "command-content.js"
+      ],
       "run_at": "document_start"
     }
   ]

package/image-size.js

@@ -0,0 +1,17 @@
+// Pixel size of a JPEG straight from its SOF marker — enough to decide whether a screenshot
+// needs downscaling without decoding it or pulling in an image dependency.
+export function jpegSize(buf) {
+  if (!buf || buf.length < 4 || buf[0] !== 0xff || buf[1] !== 0xd8) return null;
+  let i = 2;
+  while (i + 9 < buf.length) {
+    if (buf[i] !== 0xff) { i++; continue; }
+    const marker = buf[i + 1];
+    if (marker === 0xff) { i++; continue; } // fill byte
+    if (marker === 0x01 || marker === 0xd8 || (marker >= 0xd0 && marker <= 0xd7)) { i += 2; continue; } // standalone markers
+    if (marker === 0xda || marker === 0xd9) return null; // scan data / EOI before any SOF
+    const isSOF = marker >= 0xc0 && marker <= 0xcf && marker !== 0xc4 && marker !== 0xc8 && marker !== 0xcc;
+    if (isSOF) return { height: buf.readUInt16BE(i + 5), width: buf.readUInt16BE(i + 7) };
+    i += 2 + buf.readUInt16BE(i + 2);
+  }
+  return null;
+}