Skip to content

Security: jmg698/Font-Lab

Security

SECURITY.md

Security Policy

Font Lab runs on a developer's machine: it reads and edits files in your project, self-hosts font bundles, and runs a local MCP server that an AI agent drives. That makes a few classes of issue worth reporting privately before they're public.

Reporting a vulnerability

Please don't open a public issue for a security problem. Instead:

Include what you found, how to reproduce it, and the impact. We'll acknowledge within a few days, keep you posted on the fix, and credit you when it ships (unless you'd rather stay anonymous).

What's in scope

Things that would genuinely surprise a user, for example:

  • Font Lab writing outside the target project, or following a path/symlink out of it.
  • The apply/undo path corrupting or failing to back up a file it edits.
  • The MCP server or pick endpoint accepting input from somewhere it shouldn't (the pick endpoint is loopback-only by default — a way to reach it off-host would qualify).
  • A crafted project, font name, or selection.json causing code execution or an unintended write.

What's not in scope

  • The fact that Font Lab edits your source files — that's the whole point, and it's backup-first and reversible by design.
  • Issues that require an already-compromised machine or a malicious local user.
  • Bugs with no security impact — those are welcome as ordinary issues.

Supported versions

Font Lab ships from main and is published to npm continuously; fixes land in the next release. Please test against the latest published version before reporting.

There aren't any published security advisories