Kedro Builder

Kedro Builder

Identity

This skill covers Kedro node, pipeline, and catalog authoring patterns with DeAcero-specific conventions. It is the primary agent for all Kedro development tasks within the pipeline capability.

Scope

• Node authoring: Pure Python functions following the (inputs: dict) → dict contract. No side effects. No framework imports inside node functions.
• Pipeline wiring: Connecting nodes via pipeline() and Pipeline() constructs. Ensuring correct dependency ordering by matching output dataset names to input dataset names.
• Catalog registration: Registering datasets in conf/base/catalog.yml with correct typed dataset classes (pandas.CSVDataset, spark.SparkDataset, polars.CSVDataset, etc.).
• Parameter configuration: Defining parameters in conf/base/parameters/<pipeline_name>.yml and referencing them correctly in pipeline definitions.

DeAcero-Specific Conventions

• Pipeline names follow <domain>_<verb>_pipeline (e.g., ventas_ingest_pipeline, inventario_transform_pipeline).
• All DeAcero data sources must be discovered via squit-data before catalog registration.
• Business rule documentation from deacero-domain must be referenced in node docstrings.

Status

This skill is a stub. Full implementation will follow ADR-0030, which defines the Kedro integration architecture for DeAcero's data platform. Until ADR-0030 is finalized, refer to the official Kedro documentation and apply DeAcero conventions as documented in deacero-domain.

Related Skills

• deacero-domain — always invoke before accessing DeAcero data sources
• squit-data — for legacy SQL object discovery and mapping
• kedroviz-specialist — for DAG validation after pipeline wiring