Skip to content

Security: BearlySleeping/aikami

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Please do not open a public issue for security vulnerabilities.

Report privately through GitHub Security Advisories, which lets us discuss and fix the issue before it's public.

Expect an initial response within a few days. This is a small project — if you haven't heard back in a week, ping in Discord without disclosing details.

Scope

In scope:

  • The hosted client at aikami.bearlysleeping.com and the hub at hub.bearlysleeping.com
  • Auth and session handling (Better Auth, D1-backed)
  • The save backup / restore path to R2
  • Anything that lets one user read or modify another user's data
  • Code execution via crafted content — save files, character data, mods, assets

Out of scope:

  • Findings that require a compromised machine or a malicious browser extension
  • Denial of service through resource exhaustion on self-hosted deployments
  • Vulnerabilities in third-party AI model weights or upstream Docker images (report those upstream)
  • Missing hardening headers with no demonstrated impact

A note on the threat model

Aikami is offline-first by design. Gameplay state — campaigns, saves, chat history — lives in a local Turso (libSQL) database on the player's device and never requires a server. API keys entered in Settings are stored client-side and sent only to the provider endpoint the user configured.

That means the highest-severity findings are typically:

  1. Anything that exfiltrates a user's BYOK API key
  2. Cross-account access in the hub (D1 / R2)
  3. Code execution from imported content

There aren't any published security advisories