Do not report suspected vulnerabilities through a public issue. Use GitHub private vulnerability reporting.

Include the affected version or commit, reproduction steps, expected and actual behavior, impact, and a minimal proof of concept when practical. Do not include live credentials or sensitive native logs.

The current baseline does not execute a backend. It has no listener, subprocess, container, VM, credential reader, runtime environment binding, persistence, or host mutation. Its public command exposes help and version metadata only.

Future backend work is security-sensitive at the point where untrusted RAES inputs meet concrete host effects. That boundary must validate published RAES models before mutation; isolate process, container, network, and filesystem effects; keep credentials out of arguments and logs; bound readiness and timeouts; provide rollback and teardown; and redact native errors before producing RAES-owned portable diagnostics.

Repository automation is also security-sensitive. Dependencies come only from the approved registry, actions use immutable commit pins, workflow permissions are least-privilege, and Sonar credentials are unavailable to fork code.

LilRAE has one maintainer and no security response SLA. Reports are reviewed as time allows. Avoid publishing exploit details until the maintainer has had a reasonable opportunity to assess and mitigate the issue.