这个仓库中的 skills
Investigate the otaru k3s home-lab cluster for unhealthy nodes, GitOps reconciliation, workloads, storage, data plane, platform, ingress/mesh, policy, and CI health; fix safe issues via GitOps PRs; escalate destructive or database work. Invoke as /self-healing for manual or scheduled investigation runs. To start or renew the 30-minute session schedule, use /self-healing-loop.
Use when starting, renewing, or extending the session-only 30-minute self-healing schedule to the platform max TTL (currently 7 days from this run). Not for one-off cluster investigation or GitOps fixes (use /self-healing). Every invoke: create a full-window /loop job, confirm, delete other matchers, run /self-healing once when possible, report timing.
Right-size otaru workloads via KRR (CPU/memory) and Prometheus ephemeral-storage metrics. Read incident comments as guardrails, apply downsizes and upsizes in one GitOps PR, verify rollout. Also covers VPA (helm-charts/vpa), a complementary automated path for steady-state workloads with no spike pattern. Invoke as /right-sizing from /self-healing when healthy (including scheduled fires from /self-healing-loop), or manually.