castai

Systematically diagnose and fix a broken Kubernetes cluster with multiple interconnected failures

Diagnose OOMKilled pods and fix resource configuration by analyzing actual usage patterns

Configure horizontal pod autoscaling, right-size resource requests, and add resilience for services under load

Diagnose runtime state with persistent DAP debugger sessions — breakpoints, expression eval, and stepping across Go, Python, TypeScript/JavaScript, and native binaries

Run the curator — consolidate the agent-created skill library via umbrella-building