MailTrack

The bundled manifest declares host_permissions over mail.yahoo.com, mail.aol.com, outlook.live.com, outlook.office.com, and outlook.office365.com, along with content-scripts running on all six email providers. The published CWS manifest summary omits all five email-service domains and lists only the publisher's own three domains. This discrepancy means the extension is operating with host access that is not disclosed in the live store listing, preventing users and reviewers from making an informed consent decision.

{  "host_permissions": [    "https://website-track.vercel.app/*",    "https://emailtrack.xyz/*",    "https://www.emailtrack.xyz/*",    "https://mail.yahoo.com/*",    "https://mail.aol.com/*",    "https://outlook.live.com/*",    "https://outlook.office.com/*",    "https://outlook.office365.com/*"  ]}

When the extension fails to detect the logged-in AOL or Outlook user's email, it falls back unconditionally to a hardcoded developer address ('[email protected]' / '[email protected]'). This address is subsequently passed as senderEmail in the /api/extension/start-track call to emailtrack.xyz. The behaviour is likely a development artifact never scrubbed before release, but it means real tracking events from undetected AOL/Outlook users are misattributed under these identities, and the code path cannot be distinguished from an intentional fallback designed to aggregate telemetry under a controlled account.

// Fallback defaultthis.currentUserEmail = '[email protected]';return this.currentUserEmail;}// Strategy for Outlook Mailif (window.location.hostname.includes('live.com') || ...) {  ...  // Fallback default  this.currentUserEmail = '[email protected]';  return this.currentUserEmail;}

The proxy replaces the page-world XHR.prototype.send and window.fetch to read the full responseText of every Gmail API call that fires while globalTargetUuid is set. Although only 16-character hex IDs are extracted and the window is narrow, any Gmail API response body (potentially containing email metadata or content fragments) is read in full before the regex filter is applied. The injection is via a web_accessible_resource script loaded into the page context, which also means it runs outside the isolated content-script sandbox.

const originalSend = XMLHttpRequest.prototype.send;XMLHttpRequest.prototype.send = function() {  const currentUuid = globalTargetUuid;  if (currentUuid) {    this._linkedTrackerUuid = currentUuid;    this.addEventListener('load', function() {      const uuid = this._linkedTrackerUuid;      if (uuid && this.responseText) {        try {          extractAndPost(this.responseText, uuid);        } catch (_) {}      }    });  }  return originalSend.apply(this, arguments);};const originalFetch = window.fetch;window.fetch = async function(...args) {  const currentUuid = globalTargetUuid;  const response = await originalFetch.apply(this, args);  if (currentUuid) {    try {      const clone = response.clone();      clone.text().then((text) => extractAndPost(text, currentUuid));    } catch (_) {}  }  return response;};