loki-label-analyzer

This skill is an expert evaluator for Grafana Loki label strategy. It audits, designs, and improves label schemas, and diagnoses why Loki queries are slow. The core problem it addresses is label cardinality: in Loki, each unique combination of label key-value pairs creates a stream, so high-cardinality labels (like pod, user_id, request_id) inflate the index, raise storage costs, and slow queries, especially when those labels are not specified in every query selector. The skill encodes structured, opinionated rules for keeping this in check.

It provides a label evaluation framework with cardinality scoring tables, access-pattern alignment questions, static-versus-dynamic label distinctions, and consistency checks (casing, value normalization, naming conventions). It emphasizes a protected-labels allowlist (service_name, deployment_environment, job) that must never be dropped because doing so breaks Grafana Cloud correlation, Application Observability, alerts, and dashboards; for those, cardinality is treated as a value problem to remediate rather than a key to remove. It defines a required audit report structure, including a mandatory verbatim professional-services disclaimer, a downstream-dependency guardrail requiring migration notes before demoting any label, and a cost-impact analysis section that carefully distinguishes stream/index/query cost from billable ingest bytes. Reference files cover the disclaimer, protected labels, cost impact, and log-line optimization.

It targets observability engineers, SREs, and Grafana Cloud customers optimizing logging pipelines and costs. It is guidance and analysis only, read-only in nature, with no destructive operations.