Skip to content

Security: JeronimoColon/tidesman-mcp

Security

SECURITY.md

Security Policy

Reporting a vulnerability

If you find a security issue in Tidesman, please report it privately to hello@tidesman.dev rather than opening a public issue. Include the Tidesman version, your macOS version, and steps to reproduce. You will get an acknowledgement, and a fix or mitigation will be prioritized.

What Tidesman guards

Tidesman lets an AI assistant drive Apple's container tool, so its threat model is first-class:

  • It starts in read-only mode; write and destructive operations require an explicit --mode.
  • It will not mount host folders into a container without the --allow-host-mounts flag.
  • Every tool call is audited (with secret-like values redacted) to a file and the macOS unified log.
  • The binary is signed with an Apple Developer ID certificate and notarized by Apple.

Apple runs each container in its own lightweight virtual machine, so commands inside a container are isolated from your Mac. The controls Tidesman adds protect against destroying your resources and against reaching out onto the host.

There aren't any published security advisories