🛡️ Watchtower - VS Code Security Scanner
Scan untrusted workspaces for malicious code, hidden threats, and supply chain attacks — before you run anything.
Watchtower automatically scans your VS Code workspace for invisible Unicode attacks, malicious tasks, compromised extensions, and dangerous AI agent configurations. Open that sketchy GitHub repo. Clone that interview take-home. Contribute to that open source project. Watchtower checks it first.
🔍 What Gets Scanned
| Category |
What Watchtower Checks |
| 👁️ Invisible Unicode |
Hidden code using Unicode tag steganography (U+E0000–U+E007F) — invisible to the human eye |
| 📋 Malicious Tasks |
tasks.json commands that run shells, download payloads, or use encoding tricks (curl, base64, certutil) |
| ⚙️ Settings Hijacking |
Custom interpreter paths pointing to custom binaries |
| 🧩 Compromised Extensions |
Extensions cross-referenced against live threat intelligence, with auto-uninstall |
| 🤖 AI Agent Exploits |
Dangerous auto-approval settings in Copilot, Claude, Cursor, and other AI coding assistants |
| 🚀 Launch Configurations |
Suspicious pre-launch tasks hidden in .vscode/launch.json |
| 🐳 Dev Container Risks |
Security misconfigurations in .devcontainer files |
And more...
🚀 Getting Started
Recommended workflow for untrusted projects:
- Open the project — VS Code will ask "Do you trust this folder?"
- Choose "No, I don't trust the authors" to stay in Restricted Mode
- Watchtower auto-scans on open — review findings in the sidebar
- Only click "Trust Folder" after verifying the results are clean
Why Restricted Mode? Watchtower detects threats — but if you trust a workspace first, VS Code may already execute malicious tasks and scripts before the scan completes. Open untrusted, scan first, trust after.
Manual scan: Ctrl+Shift+P → Watchtower: Scan Workspace
⚙️ Configuration
Access settings via Settings → Extensions → Watchtower for global settings or the sidebar settings panel for Workspace specific settings.
Global Settings
| Setting |
Default |
Description |
watchtower.startupScans |
OnEveryProject |
OnEveryProject · OnUntrusted · Off |
watchtower.inlineFindings |
invisible |
Inline highlights: all · invisible · none |
watchtower.autoUninstallMalicious |
true |
Auto-remove extensions flagged by threat intel |
watchtower.excludedFolders |
node_modules, .git, .venv… |
Glob patterns to skip during scanning |
Workspace Settings
🎯 Real-World Attack Context
Watchtower was built in response to documented, active attack campaigns targeting developers:
🔐 Privacy
- Runs locally — no telemetry, no code is sent anywhere
- Anonymous threat intel — extension checks use an external API with no identifying data
- Open source — audit the code yourself
🤝 Contributing
Found a threat pattern Watchtower should detect? Open an issue or PR on GitHub.
📝 License
MIT — see LICENSE.md
