Agent Mandate is an early-stage, personal project. It uses only synthetic data and must not be connected to real mailboxes, accounts or credentials. There are no supported production releases.

Please report suspected vulnerabilities privately through GitHub's "Report a vulnerability" feature on this repository (Security tab), rather than a public issue. Include the affected component, steps to reproduce, and the impact you expect. You can expect an acknowledgement within a reasonable time; this is a one-person project, so there is no formal service level.

In scope: authorisation bypass in the check logic, impersonation across agents or tenants, injection through ids or text, secret exposure, and anything that lets an agent grant itself authority.

Out of scope: stopping prompt injection in general (the project controls authority, not model behaviour), compromise of the host or database, and issues that need access to the admin key. See docs/security-privacy.md for the threat model.

Never commit API keys. .env is git-ignored; .env.example lists variable names only. If a key is committed by mistake, rotate it immediately and remove it from history.