Skip to content

Security: RedLynx101/Blockwild

Security

SECURITY.md

Security policy

Supported version

Security fixes target the current main branch and the production build deployed from it. Older browser-local builds are not separately maintained.

Reporting a vulnerability

Do not open a public issue containing an exploit, private save, invite code, API key, voice credential, or player/session data. Use GitHub's private security advisory form. Include:

  • the affected route, module, or version;
  • a minimal reproduction using synthetic data;
  • the expected and observed authority boundary;
  • practical impact;
  • any safe mitigation you already tested.

Do not test against another person's world, account, or multiplayer session without explicit permission. Avoid destructive or high-volume testing against the public deployment.

Current trust boundaries

  • World and character saves are browser-local and should be treated as untrusted input during import.
  • Multiplayer is host-authoritative; guest actions must pass bounded validation before mutating shared state.
  • In-world chat is untrusted data and must not directly steer an AI runner.
  • AI companion capabilities are explicitly granted and revocable by the human host.
  • ElevenLabs and other provider credentials belong only in the external runner and must never be shipped to client JavaScript.
  • Static wiki and media assets must not become an execution path for user-supplied markup.

There aren't any published security advisories