GConnect is actively maintained on branch v2: the latest commit on v2 is the supported release. Please make sure you are running a freshly built binary are running a freshly built binary (or the bundled installer from the repo root) before reporting an issue.

Please do not open a public issue for security problems.

Use GitHub's private vulnerability reporting:

1. Go to the Security tab of this repository
2. Click Report a vulnerability
3. Describe the issue, how to reproduce it, and the impact you believe it has

You will get an acknowledgment within 72 hours and a triage decision (accepted / declined / needs more info) within 7 days. Accepted issues are fixed on the default branch and disclosed in the release notes — with credit if you want it.

In scope (things this project itself is responsible for):

• backend/auth.py — token auth, origin checks, the Tauri CORS grant
• backend/file_manager.py — path traversal, atomic writes, size caps
• backend/pty_bridge.py — the terminal WebSocket and its control channel
• backend/cluster_manager.py — remote node registry and token storage
• src-tauri/src/main.rs — the desktop shell: sidecar lifecycle, IPC commands, connection config handling
• Anything that lets an unauthenticated or cross-origin caller reach the API

Out of scope:

• Attacks that require an account on, or physical access to, the GPU host
• Containers escaping Docker (report those to Docker / the NVIDIA toolkit)
• Vulnerabilities in vendored third-party libraries (Monaco, xterm) — report upstream, but a heads-up here is appreciated
• Denial-of-service against a daemon you are authorized to use anyway

GConnect is designed to be exposed either to loopback only (default) or to a private overlay network (Tailscale/WireGuard) with a bearer token. The daemon speaks plain HTTP by design — putting it directly on the hostile internet is not a supported configuration. Please read the Security Model in the README and the full audit before deciding how much of that applies to your setup.