What problem does it solve? Keeping the Revenue Cloud ERD (erd-data.json) aligned with the actual Salesforce platform schema is error-prone: releases add and retire fields, PDF-derived artifacts pollute the data, and org differences make it hard to tell a removed field from a feature-gated one. This Skill provides the validated workflow, scripts, and classification rules to keep the ERD accurate so AI agents grounded on it stay correct. ## Core Features & Use Cases - Release refresh workflow: Extract schemas from scratch orgs, diff against the previous release's committed snapshot, patch additions, and deliberately remove retired fields. - Orphan field classification: Classify fields missing from orgs as feature-gated RC fields, other-cloud fields, or PDF artifacts, with safe-only removal and mandatory backups. - Core UDD verification: Cross-check field ownership against canonical Core UDD entity XML via codesearch before bulk removals. - Use Case: After a 262 → 264 release upgrade, build two scratch orgs, extract and diff their schemas, patch the ERD with 70 new fields, remove the 8 retired fields, regenerate the HTML ERD, and re-validate to zero gaps. ## Quick Start Ask the agent to refresh the Revenue Cloud ERD for the new release by extracting schemas from two scratch orgs, diffing against the previous release snapshot, and re-validating until gaps and missing fields reach zero.