Agent-assisted
The agent inspects, plans, reconciles authorized changes, and returns exact acceptance evidence.
Use $toad-operator in agent-assisted mode.
Prepare my one-node deployment.
Stop before billable resources.Agent-friendly, human-authorized
The open TOAD Operator skill gives coding agents a precise operating contract. It accelerates the work without inventing a second, opaque automation path.
Open and reusable
The agent inspects, plans, reconciles authorized changes, and returns exact acceptance evidence.
Use $toad-operator in agent-assisted mode.
Prepare my one-node deployment.
Stop before billable resources.The agent becomes a teacher: it explains one concept and command, then waits while you type and interpret it.
Use $toad-operator in manual-learning mode.
Do not execute deployment commands for me.
Teach one step, let me type it, then review my output.The complete AGPL-licensed skill lives in skills/toad-operator beside the infrastructure it operates.
$ pnpm run doctor -- --json
{ "ok": true, "passed": 18, "failed": 0 }Illustrative handoff · commands stay visible · actions follow your authorization
Doctor, status, verify, and diagnose produce structured evidence without changing the cluster.
Heat, manifests, and configuration remain the source of truth. Manual console repair is not the normal path.
Destroy, restore, CA rotation, and billable expansion require explicit authority.
pnpm run support
Created …/toad-support-….json
Doctor: 17 passed, 1 warning, 0 failed
pnpm run app -- diagnose my-service --json
{
"ok": false,
"services": [{ "running": 1, "desired": 2 }],
"checks": [{ "name": "https", "actualStatus": 503 }],
"hints": ["Inspect the failed task before retrying"]
}The allow-listed support report excludes credentials, keys, cloud identifiers, addresses, and logs. After the user reviews it, an agent can combine that bounded context with an application diagnosis without receiving unrestricted cloud access.