Reeve proxies API traffic, stores telemetry, and can be configured to capture message content. If you find a way to make it leak, bypass its privacy tier, or hurt the machine it runs on, I want to know privately first.

Use GitHub's private vulnerability reporting on this repository (Security tab, "Report a vulnerability"). You will get a response within a week, usually much faster. Give me a way to reproduce the problem, and if it involves a specific agent or traffic shape, the closer to a working example the better.

No GitHub account? Mail danbitengo@gmail.com with the same details.

Please do not open a public issue for a vulnerability before we have had a chance to fix it.

Things I consider security bugs: the proxy leaking request or response content anywhere the privacy tier says it should not go, the secret scanner writing an actual secret instead of its fingerprint, tier 1 storing message content, the control channel accepting commands from off the machine, and anything that lets a watched agent break out of a kill.

Things that are not: the loopback ports being open on localhost (deliberate), and vulnerabilities in the agents Reeve watches, which are yours.

"Loopback" comes with two caveats. The control channel and proxy bind 127.0.0.1 and stay there, but the OTLP receiver, though it also defaults to loopback, can be rebound by REEVE_ADDR; a non-loopback address there opens an unauthenticated ingestion port, so only set one behind a network you trust. And loopback is not authentication: any process on the machine can inject spans or reach the control channel, so Reeve trusts every local process and should not run on a shared host with people you don't trust.

Pre-1.0, only the latest release gets fixes. From 1.0 on, the latest minor of the current major.