core-agent

Walk a user through configuring an unattended core-agent — single-agent monitor or multi-agent team. Use when the user asks to "set up an autonomous agent", "run core-agent unattended", "configure a monitor", "build a multi-agent team", "deploy core-agent to production", "set up a long-running agent", asks how to bound budgets, asks about crash-resume, or describes a goal that obviously needs unattended execution (e.g., "watch the deploy", "monitor SLO breaches", "scan the fleet every hour").

Walk a Go developer through embedding core-agent in their own binary. Use when the user asks "how do I embed core-agent", "use core-agent as a library", "build my own agent on core-agent", "agent.New", "custom prompter", "custom tools", "custom provider", "HTTP-served agent", "integrate core-agent into [Go project]", or asks any question implying they want to use the Go API rather than the bundled CLI.

Handle a Kubernetes event inject shaped like {"kind": "k8s-event", "reason": "<Reason>", "namespace": "...", ...}. Loads the reason-specific reference and drives the diagnose → fix → verify loop for any k8s failure mode. Falls back to a generic playbook (references/_fallback.md) for unknown reasons.

Walk a user through configuring core-agent for interactive (TUI / REPL) use against their project. Use when the user asks to "set up core-agent", "configure core-agent for my repo", "help me get started with core-agent", "what should my AGENTS.md look like", "what permissions should I use", "how do I add a skill", or asks any question implying they're customizing core-agent for the first time or expanding an existing configuration.