Please do not report security vulnerabilities through public GitHub issues, discussions, or pull requests.
Instead, report them privately via GitHub's private vulnerability reporting:
Go to the repository's Security tab → Report a vulnerability.
Please include, as far as you can:
- a description of the vulnerability and its impact,
- the affected version (see
Settings → About) and platform, - step-by-step instructions to reproduce it, and
- any proof-of-concept or logs.
We'll acknowledge your report as quickly as we can, keep you updated on the fix, and credit you in the release notes if you'd like.
A few Lume features are inherently security-sensitive — reports about them are especially welcome:
- Remote control — the HTTP/WebSocket server that mirrors panes to another
device: QR pairing (one-time secret in the URL fragment), per-device keys and
revocation, the end-to-end encrypted channel (NaCl secretbox, see
src-tauri/src/remote_proto.rs), the served web page, the Lume ↔ Lume client and the optional cloudflared tunnel. Known limit: the page itself is served over plain HTTP on the LAN, so an active attacker on that network could serve a modified page; the encryption protects against passive sniffing and against the tunnel provider. - AI providers — handling of API keys and the commands spawned for CLI providers.
- Shell integration & PTY — anything that could lead to unintended command execution.
Lume is pre-1.0 and ships from a single main line. Security fixes land in the
latest release; please make sure you're up to date (the app auto-updates)
before reporting.