copilot-howto

Use when a pod is crashing, not starting, or behaving weirdly on Kubernetes (EKS, GKE, AKS, or vanilla). Triggers on: CrashLoopBackOff, OOMKilled, ImagePullBackOff, ErrImagePull, Pending, ContainerCreating, readiness probe failing, liveness probe failing, pod stuck terminating, node not ready, kubectl describe events, "why is my pod crashing", "pod won't start".

Use when the user asks to deploy, upgrade, roll back, or release a service to Kubernetes. Also triggers for questions about Helm chart values, image tags, release history, rollout status, canaries, or blue/green cutovers. Relevant phrases include: "deploy to staging", "bump the image", "rollback", "helm upgrade", "release", "promote to prod", "what image is in production".

Use when an incident is in progress or was just resolved. Triggers on: production incident, alert firing, pager, on-call, outage, p0, p1, sev1, sev2, postmortem, retro, customers reporting, elevated error rate, "something is down", "we have an incident".

Use when the user wants to plan, apply, or change infrastructure with Terraform or OpenTofu. Triggers on: tofu plan, terraform plan, tofu apply, terraform apply, tf state, "change the infrastructure", "update the Terraform", "add a resource", "import existing resource", "drift", "state lock", "workspace".