VS Code has no permission model. Any extension you install runs as ordinary Node.js with the full reach of your account: your files, your SSH keys, your terminal. Nothing had to ask, so nothing prompted you. ExtGuard audits the ones you already have, against 1,981 known-malicious records and 1,300 blocked publishers, for credentials hardcoded into shipped code, and for what each can reach. Runs entirely on your machine. No account, no telemetry, no network calls while scanning. Free and MIT.
I built this after noticing something obvious I had never actually thought
about: I had 81 extensions installed and had read the source of none of them.
VS Code has no permission model. No prompt, no sandbox, no list of what an
extension asked for, because it never had to ask.
The interesting part was not detection, it was restraint. My first version
flagged 57 of my 82 extensions. That is not a security tool, that is a tool you
learn to ignore, and then you ignore the one finding that mattered. Today it
flags zero on the same machine and I trust it more for that.
It will not catch a competent attacker. Pattern matching never does. It catches
the careless majority: typosquats, credentials left in shipped code, and
extensions pulled from the marketplace that are still sitting on your disk.
Free, MIT, runs locally. If it flags something of yours incorrectly, send me the
extension ID and I will fix it.