Skip to content

Security: fdkevin0/mob

SECURITY.md

Security policy

Mob holds model provider keys, reads your mailbox, runs commands in a sandbox and can be reached remotely, so we take reports about it seriously.

Reporting a vulnerability

Report privately through GitHub's private vulnerability reporting. Please do not open a public issue, pull request or discussion for a vulnerability.

Include what you can of:

  • the affected part (see Scope) and version (mob --version, or the date you used usemob.app);
  • steps to reproduce, or a proof of concept;
  • what an attacker gains and what they need first (for example, a local account, a malicious MCP server, or a crafted email).

Mob is maintained by one person. You should get an answer within 7 days; we agree on a fix and a disclosure date with you, and credit you in the advisory unless you would rather not be named.

Supported versions

Version Supported
The latest stable release (npm install -g usemob@latest) Yes
The latest nightly (usemob@nightly) Yes
Older releases No; upgrade first

Scope

In scope:

  • the mob program and its npm packages (usemob, @usemob/*), the release archives and the desktop app;
  • the local service: its HTTP API, access tokens, pairing, approvals and remembered grants, the sandbox, and the plugins and channels in this repository;
  • usemob.app: the web app, sign-in, the account page, the encrypted runner relay and the Workers AI route.

Out of scope:

  • model providers, MCP servers, skills and coding agents that you add yourself, except where Mob fails to enforce an approval, a sandbox limit or another protection it documents;
  • what a model chooses to say or do within the tools you approved;
  • attacks that need an account on your machine with access to Mob's data directory.

How approvals, the sandbox and remote access are meant to behave is described in tools and approvals, sandbox and remote access.

There aren't any published security advisories