Convert a migration request into a structured, zero-downtime plan with tested rollback procedures.
Use this skill when the user asks to:
Classify the migration type: database | service | infrastructure | hybrid.
Load patterns for that type from {baseDir}/references/migration_patterns_catalog.md.
Run compatibility check if schemas or APIs are provided: bash
python {baseDir}/scripts/compatibility_checker.py --before=<old> --after=<new>
Identify dominant zero-downtime technique (expand-contract, dual-write, CDC, strangler-fig, blue-green, canary). Load {baseDir}/references/zero_downtime_techniques.md if the best technique is unclear.
Generate migration plan with phased execution: bash
python {baseDir}/scripts/migration_planner.py --config=<config.json>
Generate rollback procedures for each phase: bash
python {baseDir}/scripts/rollback_generator.py --plan=<plan.json>
Define validation checkpoints: row counts, checksums, business-logic queries. Load {baseDir}/references/data_reconciliation_strategies.md for reconciliation patterns.
Emit output: phased plan + rollback procedures + runbook checklists + success metrics.
Functor Information Loss (or equivalent: Irreversible Data Loss Risk) when a phase cannot be rolled back.TierCriteriaRequired before executionLOWadditive schema changes, no data transformationstaging validationMEDIUMdata transformation, dual-write window, service cutoverstaging + load test + rollback drillHIGHdestructive changes, cross-cloud, compliance scopeall above + stakeholder sign-off
Information Loss).references/ instead.Before emitting the plan, verify: