What problem does it solve?
When migrating a Kubernetes Config Connector resource controller from the legacy Terraform/DCL approach to the Direct approach, the new controller often detects spurious differences during takeover and issues unwanted writes to GCP. This Skill provides a mandatory, ordered workflow to diagnose, explain, and eliminate those takeover diffs so the Direct controller adopts resources with a clean 0-write no-op reconciliation.
Core Features & Use Cases
- Structured 4-Step Workflow: Enforces diagnosing on real GCP first (via
./hack/record-gcp), identifying root causes, fixing comparison logic in the Direct controller, and validating against real GCP.
- Root Cause Patterns: Covers derived/computed fields, casing and alias mismatches, default value discrepancies, legacy diff suppression, and URL vs. relative path normalization.
- Migration Journals: Includes per-resource journals (BigQueryDataset, ComputeNodeTemplate, ComputeTargetHTTPSProxy, NetworkServicesHTTPRoute) documenting real takeover diff symptoms and their resolutions.
- Use Case: While migrating
ComputeNodeTemplate to Direct, the takeover diff shows cpu_overcommit_type defaulting to NONE on GCP but nil in KRM; the Skill guides adding default normalization in the compare function and re-recording real GCP logs to verify a 0-write takeover.
Quick Start
Ask the agent to run the migration diff workflow for a specific fixture, starting with recording real GCP traffic for TestMigrationToDirect and analyzing the generated _migration_diffs.json file.