data-analytics-routing

# Data workflow selection 1. Identify the user's decision, required evidence, available sources, and expected artifact. A task that only asks for a text correction does not require a data workflow. 2. Verify source access and key definitions. Use supplied files, authorized databases, connectors, or documented APIs; do not assume a particular plugin manifest or bundled demo dataset. 3. Select the narrowest matching workflow: | Need | Workflow | | --- | --- | | Trustworthiness, missingness, grain, joins | Data-quality assessment | | Explain a metric change | Metric diagnosis | | Define measures and targets | KPI design | | Make a business or product decision | Evidence-backed analysis | | Communicate findings | Report or dashboard | | Explain quantitative relationships | Visualization | 4. Locate the corresponding skill in the current library if present. Use host capabilities that actually exist. Do not require a fixed connector, application, query engine, or provider-specific tool namespace. 5. Keep methods and sources inspectable. State limits when required evidence is unavailable instead of inventing results or silently substituting a weaker source. 6. Deliver in an available format. Analysis does not authorize publishing, messaging, or live data changes.