tracing-downstream-lineage_skill

Overview

This skill traces downstream data lineage and performs impact analysis to answer “what breaks if I change this?”. It helps you discover direct consumers, map full dependency trees, and quantify the blast radius before modifying tables or DAGs. Use it to produce a clear impact report with owners, criticality, and recommended mitigations.

How this skill works

The skill inspects DAG source code, table/view metadata, and BI/dashboard connections to find anything that reads from a target table or is produced by a target DAG. It recursively builds a dependency tree, categorizes downstream assets by criticality, and evaluates risks for schema, data, timing, or deletion changes. The output includes a summary, ASCII impact diagram, detailed impacts table, risk assessment, and recommended actions.

When to use it

• Before changing a production table schema or data format
• Prior to deprecating or deleting a dataset or DAG
• When assessing risk for column renames or removals
• Before scheduling major ETL timing or volume changes
• When preparing a stakeholder notification or migration plan

Best practices

• Start by locating direct consumers in DAG code and view definitions before exploring BI tools
• Treat production dashboards, customer-facing apps, and ML models as highest criticality
• Categorize downstream assets (Critical/High/Medium/Low) to prioritize mitigation work
• Validate owners for each downstream asset and notify them early
• Run targeted end-to-end tests for schema and data changes and schedule changes in a low-impact window

Example use cases

• Change a table column name: identify affected DAGs, dashboards, and ML features and list required updates
• Evaluate deleting a legacy table: map all consumers and required migrations before deprecation
• Assess an ETL timing shift: find dashboards and reports that expect specific update windows and flag SLA impacts
• Prepare a release plan: generate an impact diagram and stakeholder checklist for a planned schema migration
• Investigate a production failure: find immediate downstream assets to inspect for cascading errors