cx-telemetry-querying

This skill is the entry point for investigating, debugging, and answering data questions using Coralogix observability data through the cx CLI. It solves the routing problem of an incident or analysis question — deciding where the relevant signal lives across the four telemetry pillars (metrics, logs, traces/spans, and RUM) and which reference files to load before querying, so an agent can move efficiently from a vague symptom to a concrete query.

The skill provides a quick routing table mapping question types to a first-choice pillar and fallback, plus a discovery workflow for ambiguous cases that searches metrics, semantic log/span fields, and the codebase in parallel. It documents read-only cx commands — cx logs, cx spans, cx metrics, cx dataprime, cx search-fields — and emphasizes that all query commands are read-only and safe to run without confirmation. Companion references cover DataPrime syntax, PromQL guidelines, and per-pillar querying (logs, spans, metrics with instant vs. range queries and label discovery, and RUM fields), along with structured investigation workflows for HTTP errors, latency, and availability, plus retry logic and summarization guidance.

It targets SREs, platform engineers, and developers who use Coralogix and want an AI agent to drive telemetry investigations. Prerequisites include the cx CLI with a configured profile carrying a Coralogix API key or OAuth; the skill prompts the user to run cx profiles add when credentials are missing.