Security policy

agent-sidecar governs what AI agents may do, so security reports matter a lot to this project.

Supported versions

The project is a v1alpha1 MVP. Only the latest commit on main gets fixes. It is not production-ready yet. See the known limitations in docs/operations.md.

Report a vulnerability

Please do not open a public issue for a security problem.

Use GitHub private vulnerability reporting: open the repository’s Security tab and choose Report a vulnerability, or go to https://github.com/f8fvwgzc/agent-sidecar/security/advisories/new.

Please include:

• what an attacker can do, for example get around an authz rule, a guardrail or a budget, or read data from the audit log;
• the charter and the requests that show it;
• the commit you tested.

You should get a first answer within 7 days. A fix and an advisory follow once the problem is confirmed.

Scope

In scope: the sidecar (attached), the runtime (attache-runtime), the CA (embassy), the CLI (attache), the charter validation and the Kubernetes manifests in deploy/.

Already known limitations listed in docs/operations.md and in CONTRACT.md section 11 are not new vulnerabilities. A report that shows a consequence the docs do not mention is welcome.