Skip to content

Security: peter-olom/manor

Security

SECURITY.md

Security Policy

Manor is a trusted personal worker appliance. It is designed for one operator on infrastructure they control, not for multi-tenant hostile-code execution.

Supported Versions

Until the project has tagged releases, security fixes target the current main branch.

Security Model

Manor assumes:

  • the Docker host is trusted by the operator
  • Butler is not exposed directly to the public internet
  • Codex and Butler run inside the same trusted appliance boundary
  • the runtime broker is trusted to manage Docker resources
  • previews and disposable services are isolated for operational hygiene, not as a complete security sandbox

Manor does not currently claim:

  • multi-tenant isolation
  • protection from a malicious local operator
  • protection from a compromised Docker daemon or host
  • safe public exposure of the Butler UI
  • safe execution of arbitrary untrusted code

Use private ingress, local binding, VPN, tailnet, or similar controls when accessing Manor remotely.

Reporting a Vulnerability

Report vulnerabilities through GitHub private vulnerability reporting.

Do not include exploit details, secrets, tokens, private URLs, or reproduction steps in a public issue.

Please include:

  • affected component
  • expected impact
  • reproduction summary
  • affected configuration
  • whether credentials, host access, or public exposure are required

Handling Secrets

Do not commit API keys, session tokens, ChatGPT auth data, GitHub credentials, local state volumes, captured browser sessions, or proof artifacts that contain sensitive user data.

If a secret is exposed, rotate it immediately. Removing it from a later commit is not enough.

Hardening Expectations

For remote use:

  • keep Butler behind private access controls
  • restrict inbound host ports
  • keep Docker and base images updated
  • review runtime logs when debugging suspicious behavior
  • prefer short-lived test credentials inside previews

Security-sensitive changes should include validation notes and, when practical, a small regression test.

There aren't any published security advisories