grandplan is an offline-first, single-user desktop tool. Its trust model — local-process trust, the optional shared-secret token for LAN intake, and unencrypted local storage — is described in the README's Security model section. Please read it first to understand what is and isn't in scope.

Please do not open a public issue for security vulnerabilities.

Report privately via GitHub's Report a vulnerability (repo Security → Advisories → Report a vulnerability). If private reporting is unavailable, email the maintainer at [email protected] with a description and, ideally, a minimal reproduction.

Expect an acknowledgement within a few days. As a personal project maintained in spare time, please allow reasonable time for a fix before any public disclosure (coordinated disclosure is appreciated).

Pre-1.0: only the latest main receives security fixes.

In scope: the core pipeline, the HTTP intake endpoint (grandplan serve / up), the folder-watch capture, the Ollama adapter, and vault writes.

Out of scope: issues that require an attacker who already has local code execution or filesystem access equivalent to the user running grandplan — the tool explicitly assumes the local user and machine are trusted. The HTTP intake has no authentication on 127.0.0.1 by default by design; a routable (LAN) bind requires a token. See the README security model for the rationale.