Core OPAL: pipeline generation, dataset kinds, column selection, routing

This skill provides core OPAL guidance for generating valid Observe OPAL pipelines. It does NOT contain domain-specific patterns for logs, spans, metrics, joins, durations, or other specialized areas — those live in opal-* references that MUST be loaded before writing any OPAL.

Workflow

1. Determine the dataset kind and interface for each data source (see Dataset Kinds below).
2. MANDATORY — STOP and load references NOW. Consult the Skill Index below. You MUST read EVERY opal-* reference file that matches the dataset kind or query intent BEFORE writing any OPAL. Read files from /skills/generate-opal/references/ (e.g., cat /skills/generate-opal/references/opal-logs.md). If multiple references match, read ALL of them. Do NOT proceed to step 3 until all matching references are loaded. Skipping this step leads to incorrect queries — references contain critical dataset-specific patterns, multi-dataset workflows, and syntax that this core skill does not cover.
3. Write the OPAL pipeline using the syntax rules in this skill combined with the loaded domain skills.

References

MANDATORY: You MUST load references before writing OPAL. Check this routing table and read every reference file that applies (e.g., cat /skills/generate-opal/references/opal-logs.md). If multiple references are listed, read ALL of them — do not skip any. Do NOT write any OPAL pipeline until all matching references are loaded.

OPAL join/lookup/set-operation syntax is non-standard — do NOT rely on SQL knowledge. If the answer requires data from more than one dataset, you MUST read opal-join-patterns before writing any OPAL.