kennel is a security tool, so reports about its isolation claims are the most valuable kind.

Please do not open a public issue. Report privately through GitHub's "Report a vulnerability" (Security advisories) on this repository. Include the engine and version, OS, your kennel version, and the smallest reproduction you can, ideally a failing kennel verify check.

We aim to acknowledge reports within 3 working days. Coordinated disclosure: we will agree a timeline with you and credit you unless you prefer otherwise.

Maintainers: private vulnerability reporting must be enabled in the repository settings before the first public release.

Release archives are checksummed, signed with Sigstore (keyless) and carry build-provenance attestations and SBOMs. See docs/releasing.md for the exact verification commands.

In scope: anything that lets an agent inside the sandbox reach the network, the host, or the gateway's control surface in a way the documentation says it cannot; anything that lets it modify protected paths or its own policy; flaws in the proxy, policy matcher or generator.

Out of scope (documented non-goals, see docs/threat-model.md once written): kernel or container runtime escapes, data sent to hosts you allowlisted, domain fronting through shared CDNs, and a compromised host.

Pre-release: only the latest commit on master.