Skip to content

Security: loitc910-maker/javis-os

Security

SECURITY.md

Security Policy

Javis runs an AI brain with full rights on the machine it is installed on, so security reports matter a great deal to us. Thank you for taking the time.

Reporting a vulnerability

Please do not open a public issue for a security problem. Report it privately instead:

  1. Open a private report at Security → Report a vulnerability on this repository. Only you and the maintainers can see it.
  2. If that page is not available, open a normal issue that says only "I have a security report, please give me a private contact", with no details, and the maintainer will reply with one.

Useful things to include: the version (shown under Settings → Updates, or in the VERSION file), how Javis is deployed (Docker, VPS, local), the steps to reproduce, and what an attacker could do.

What to expect

  • An acknowledgement within a few days.
  • A fix released as a new version as soon as it is ready. Javis updates itself through the Updates page, so fixes reach users quickly.
  • Credit in the changelog if you would like it.

Supported versions

Only the latest release receives security fixes. Please check that the problem still exists on the current main before reporting.

Scope notes

Some behaviour is by design and documented: the CLI brains can run shell commands on the host, user plugins run real Python inside the server (loaded by default once enabled in their manifest; JAVIS_ENABLE_USER_PLUGINS=false turns them off), and connections at Full access can act on real accounts. A way to get past those safeguards (login, CSRF and DNS-rebinding protection, the plugin gate, connection permission levels) is exactly what we want to hear about.

There aren't any published security advisories