Relay is remote code execution as a service by design: the Relay dispatches administrative Operations to relay-daemons on Nodes. Vulnerability reports are the highest-priority work in this repository.

Relay is alpha. Only main is supported; fixes land on main and ship in the next release. docs/operations/release-artifacts.md describes what a release is.

Report privately through GitHub private vulnerability reporting: https://github.com/Felag-Engineering/relay/security/advisories/new (or the repository's Security tab, then "Report a vulnerability"). This is the only channel. Do not open a public issue, discussion or pull request for a suspected vulnerability.

A useful report contains:

• the affected component: the Relay, the relay-daemon, or a packaged unit;
• the version or commit (relay --version / relay-daemon --version);
• steps to reproduce, and the impact;
• the architecture invariant (numbered, in AGENTS.md) or trust boundary it breaks, if you know.

Please do not include live credentials, private keys, or Enrollment Tokens.

We aim to acknowledge a report within 5 business days and give an initial assessment after triage. The report is worked in a draft GitHub Security Advisory with the reporter as a collaborator. The fix is developed in the advisory's private fork and merged to main, and released, when the advisory is published. The reporter is credited unless they decline. Please hold public disclosure until then.

In scope:

• the relay binary (the Relay) and the relay-daemon binary;
• the packaged units: the Relay's systemd unit and example relay.toml (relay/packaging/), and the relay-daemon's systemd unit, example EnvironmentFile and shipped Template catalog (daemon/packaging/).

Out of scope:

• the test and demo fleets (deploy/test/, deploy/demo/, docker/ and its fixtures, make dev-fleet-up, make demo-up, make demo-up-oauth): local development fixtures with dev-only credentials, not a deployment;
• vulnerabilities in third-party dependencies with no demonstrated path through Relay; report those upstream.

docs/security/threat-model.md maps the trust boundaries, the adversaries, and which invariant defends each. A report showing one of the fourteen architecture invariants in AGENTS.md failing is in scope. A gap already in the Known gaps table is tracked; cite its issue rather than reporting it again.