Skip to content

Security: philipnee/mvmt

SECURITY.md

Security Policy

mvmt exposes personal local data. Security bugs should be handled privately and conservatively.

Supported Versions

mvmt is pre-1.0. Security fixes target the latest code on main until a stable release process exists.

Version Supported
main yes
< 0.1.0 no

Reporting A Vulnerability

Please do not open a public issue with exploit details.

Use GitHub private vulnerability reporting for this repository if it is enabled. If private reporting is not available, open a minimal public issue asking for a private security contact and do not include sensitive details.

Include:

  • Affected version or commit.
  • Operating system.
  • Whether mvmt was running local-only or through a tunnel.
  • Minimal reproduction steps.
  • Impact: what data or capability became accessible.
  • Whether the issue requires local user access, local network access, or public internet access.

Security Model Summary

Current protections:

  • HTTP mode binds to 127.0.0.1 only.
  • HTTP /mcp and /health requests require a bearer token.
  • Origin checks reject non-localhost browser origins unless allowlisted.
  • Local folder access is scoped by explicit mounts.
  • Scoped API tokens can restrict each client by virtual path and action.
  • Write/remove access is opt-in at both the client policy and mount level.
  • Protected paths block write/remove even when write access is otherwise allowed.
  • Tunnel mode is closed to legacy all-mount access unless explicitly overridden.
  • Auth, MCP, and health routes have rate limits.
  • Stdio child processes receive a scrubbed environment.
  • Tool calls are written to a local audit log.

Known limits:

  • Tunnel mode exposes the configured HTTP surface to the internet. Use narrow mounts and scoped API tokens.
  • Localhost traffic is plaintext.
  • protect blocks write/remove, not reads. Use exclude to hide readable secrets from listing, indexing, and reads.
  • The audit log can include truncated argument values.
  • OAuth/tunnel auth is pre-1.0 and should be treated conservatively.

See docs/threat-model.md for the current threat model and docs/security-memo.md for implementation notes.

There aren’t any published security advisories