What problem does it solve? Answering observability questions about Nebius cloud resources requires navigating Grafana datasources, PromQL, LogQL, and TraceQL while respecting strict scope, authorization, and read-only boundaries. This Skill provides a bounded, fail-closed workflow for querying metrics, logs, traces, and dashboards through a pre-configured grafana-nebius MCP server without risking mutations, credential misuse, or unscoped queries. ## Core Features & Use Cases - Scoped read-only queries: Resolve explicit tenant, project, and resource scopes into fail-closed predicates, discover datasources and labels, and run bounded Prometheus, Loki, and Tempo queries with fixed execution limits. - Structured result reports: Return ranked, signal-aware reports with per-group statistics, coverage and access evidence, and explicit truncation disclosure instead of raw telemetry dumps. - Embedded evidence provider: Supply troubleshoot and sdlc-evaluate workflows with a fixed facts envelope, one-time readiness checks, and cumulative fast/deep query budgets. - Use Case: Ask for CPU and memory usage per node in a specific MK8s cluster over the last 24 hours and receive a problem-first ranked table with minimum, average, and maximum values, plus coverage and limitation notes. ## Quick Start Ask the agent to show GPU utilization by node for a specific project ID over the last 24 hours using the configured Grafana MCP server.
