KubeSense SLOs

An SLO here is a stored definition plus a background evaluator. Every evaluation_interval_seconds the evaluator scores one bucket, writes it to ClickHouse, and emits Prometheus series (kubesense_slo_*). Compliance is then a rolling aggregate over evaluation_window_days.

[!IMPORTANT] There are no SLO MCP tools. Logs, traces, metrics and alerts have them; SLOs do not. Everything here is the REST API at /api/slo, authenticated exactly like the MCP server (x-api-key, or Authorization: Bearer <token>). Use the MCP query tools to discover metric and field names, then build the SLO body.

So this skill needs two things an MCP connection does not give you: a way to make HTTP calls (a shell with curl, or an HTTP tool), and the API credentials as a value you can put in a header — the MCP server's own credentials are not readable from inside a tool call. An agent with MCP alone can still build and explain the payload, and should say plainly that the user has to send it.