TokenDiet runs inside your own editor, on your own machine, against your own code. There is no TokenDiet server and no account — the plugin is just the Markdown files in this repo, loaded by your own agent. That keeps the attack surface small, but a malicious change to these files (or a malicious fork) could still instruct an agent to do something harmful. We take that seriously.

Please do not open a public issue for a security problem. A public report tells attackers about the issue before a fix is out.

Instead, report it privately:

• Use GitHub's Report a vulnerability (Security tab → "Report a vulnerability"), or
• Email security@fortytude.dev with the details.

Please include:

• What the issue is and where (file and, if relevant, the exact instruction).
• How to reproduce it or a proof of concept.
• The impact you think it has.

Because TokenDiet is a set of agent instructions, the security concerns that matter most here are:

• Prompt-injection or unsafe instructions — any wording in a skill, command, or reference file that could steer an agent into exfiltrating code, running a destructive command, or applying an unsafe edit.
• A fix the plugin offers to apply that could corrupt code or leak data.
• Impersonation — a fork or lookalike marketplace passing itself off as the official TokenDiet. The official source is Fortytude/TokenDiet; anything else is not us.

We'll acknowledge a report as quickly as we can, keep you posted while we work on a fix, and credit you when it ships (unless you'd rather stay anonymous). Since distribution is a git repo, the fix reaches users on their next plugin update.