Skip to content

Security: santoshakil/ai_debug

SECURITY.md

Security

ai_debug runs inside the app and can expose sensitive runtime data. Release startup is blocked unless allowRelease: true is set explicitly. Keep it in debug or controlled internal builds.

  • loopback is the default and recommended bind address
  • loopback prevents LAN access but is not app-private; another local process may connect, so authentication is always required
  • every config has an explicit or randomly generated access token
  • keep fixed tokens in an ignored local environment file and out of logs, source control, screenshots, and shared reports
  • a non-loopback bind requires explicit unsafeLan: true and uses plain HTTP
  • /healthz and the static dashboard shell are public; debug data and commands require the token
  • file access requires explicit absolute roots
  • use read-only, stable file roots; writable roots can change during a request
  • app state and network requests require explicit instrumentation
  • automation, runtime controls, outbound probes, telemetry, VM service access, and mDNS are opt-in
  • outbound probes can resolve and connect to private addresses; enable them only for a trusted local client
  • telemetry accepts HTTPS collectors and loopback HTTP only
  • widget, semantics, error, log, route, and state data can contain private app content even when a tool is read-only
  • do not register secrets, tokens, private headers, raw request bodies, or personal data

Report security issues through a private GitHub security advisory instead of opening a public issue.

There aren't any published security advisories