SchemaShield — Evidence-First Schema Change Preflight by Kai Relaunch DeptSchemaShield — Evidence-First Schema Change Preflight by Kai Relaunch Dept

SchemaShield — Evidence-First Schema Change Preflight

Kai Relaunch Dept

Kai Relaunch Dept

SchemaShield — Evidence-First Schema Change Preflight
A schema change can look harmless in a pull request while breaking models and dashboards one or two hops downstream. RELAUNCH DEPT. built SchemaShield to turn catalog context into a reviewable risk decision before merge.
The public replay evaluates three deterministic schema-change scenarios and generates seven review artifacts: compatibility SQL, a dbt schema patch, an impact report, a pull-request summary, provenance, submission status, and a writeback plan. The rename scenario is classified HIGH and identifies exactly two downstream datasets.
A separate verified local round trip used DataHub OSS v1.6.0: Agent Context Kit read schema fields and performed approval-gated tag and document writebacks; the DataHub Python SDK read two-hop lineage. The hosted replay uses only synthetic data and performs no live writebacks.
AI disclosure: AI-assisted development and copy. Public claims were checked against reproducible synthetic evidence; no production or private data is used.
Like this project

Posted Aug 6, 2026

Preflights risky schema changes, exposes two-hop impact, and generates seven review artifacts before merge.