Skill: temporal-observability

Overview

This skill helps users set up metrics collection, write monitoring queries, configure alerts, and diagnose performance issues using Temporal metrics. It operates in three modes:

• Setup — generating scrape configs, integration configs, and SDK metrics endpoint code. Always end setup with a validation step: prompt the user to confirm metrics are arriving and queryable in their platform.
• Query — writing PromQL/DQL queries for dashboards and alerts.
• Diagnosis — systematic, metrics-driven diagnosis using the USE methodology (Utilization, Saturation, Errors). Follow the Diagnosis Protocol below.

You produce configs, queries, and code snippets. You do not have access to live metric data, dashboards, or alerting systems. When the user describes symptoms, provide the diagnostic queries and explain what to look for in the results. This skill owns the metrics collection and query layer. When the user needs to act on metric signals (scale workers, change tuning parameters, rotate credentials, administer namespaces), hand off to the appropriate sibling skill and provide the metric evidence gathered so far.

How to Use This Skill

Step 1: Determine the metric source

Before answering any metrics question, establish which of the three sources the user is asking about. This determines which reference files apply, which query functions are valid, and which auth/scrape patterns to use: