Open-source tool that simulates what coding interview platforms detect.
Test your setup before the real thing.
Built by blind.codes — the invisible AI assistant for coding interviews.
👉 blindcodes.github.io/stealth-benchmark
| Check | Risk | Method |
|---|---|---|
| Window focus lost | 🔴 High | blur / focus events |
| Tab switch | 🔴 High | Page Visibility API |
| Global shortcut (Cmd+Shift+[key]) | 🔴 High | keydown event |
| DevTools open | 🟡 Medium | Window size delta + console getter heuristic |
| Multiple screens | 🟡 Medium | window.screen.isExtended |
| Clipboard access granted | 🟡 Medium | Permissions API |
| Screen share availability | 🔵 Low | navigator.mediaDevices.getDisplayMedia |
| User agent / OS | 🔵 Low | navigator.userAgent |
| Timezone & locale | 🔵 Low | Intl.DateTimeFormat |
- Running processes — requires a native app
- Specific overlay windows — not accessible from the web
- Kernel-level hooks
Just open index.html in a browser. No build step, no dependencies.
git clone https://github.com/blindCodes/stealth-benchmark
cd stealth-benchmark
open index.htmlPRs welcome. If you know of a detection technique used by interview platforms that isn't covered, open an issue or submit a PR.
Each check follows this interface:
{
id: 'my-check', // unique identifier
name: 'My Check', // display name
desc: 'What it does', // shown under the name
risk: 'high', // 'high' | 'medium' | 'low'
run: async () => ({
pass: true, // true = safe, false = flagged, null = info only
detail: 'message' // shown in the event log
})
}MIT