data-intake-router

# Data Intake Router Agent routes data and Fabric work to one primary data skill. ## When to Use | User prompt | Use | |---|---| | User asks for Fabric, KQL, ingestion, pipeline, semantic model, or data-quality work and the exact concern is unclear | Use this skill | | Request spans more than one data-oriented specialist | Use this skill | | User asks which data skill owns the work | Use this skill | ## When Not to Use | User prompt | Route | |---|---| | Request is already narrowed to one data specialist | Use that specialist | | Request is application-domain design and data tooling is only a supporting detail | Use the domain skill first | ## Routing Rules | Topic | Rule | |---|---| | Entry ownership | Broad data and Fabric routing belongs here | | Fabric split | KQL, ingestion, pipelines, and semantic models stay separated unless the request explicitly combines them | | Quality split | Data-quality work loads lineage or contract specialists only when the primary concern requires them | | Specialist loading | Choose one primary data skill first, then only concrete secondaries | ## Verification - [ ] Agent chose one primary data skill - [ ] Agent kept Fabric routing specific to the observed failure or authoring mode - [ ] Agent added a secondary data skill only when the primary scenario required it ## Routing Table | Intent | Primary skill | Optional secondary | |---|---|---| | KQL failure or performance diagnosis | `fabric-kql-diagnostics` | none | | Lakehouse ingestion shape or file landing | `fabric-lakehouse-ingestion` | `data-quality-reporting` | | Fabric pipeline failure or orchestration issue | `fabric-pipeline-diagnostics` | none | | Semantic model authoring | `fabric-semantic-model-builder` | `data-contract-frontmatter` | | Data-quality reporting or rule evaluation | `data-quality-reporting` | `data-lineage` |