← ClaudeAtlas

postgres-migration-safetylisted

Plan and review PostgreSQL production changes for locks, scans, rewrites, compatibility, bounded backfills, constraint validation, concurrent indexes, replication, observability, failure cleanup, and rollback or roll-forward. Use when a target schema or data change is approved for rollout. Do not use to choose the domain model, diagnose a query, or design the backup program.
genaptic/skillsets · ★ 1 · API & Backend · score 70
Install: claude install-skill genaptic/skillsets
# Outcome Produce a concrete, reviewable result for the workflow below without overstating what was observed, executed, or verified. ## Compatibility Portable across Claude Code, Codex, and OpenCode. Guidance targets PostgreSQL 18, but verify lock, rewrite, transaction, partition, and provider behavior on the exact deployed version. The optional Python 3.11 auditor reads SQL text only and cannot prove safety. ## Use this skill when - Approved DDL or data changes must be sequenced for a live PostgreSQL system. - A large table needs columns, types, constraints, indexes, backfills, repartitioning, or destructive cleanup. - Migration-runner transactions, lock waits, long snapshots, WAL, replicas, mixed application versions, or failure state create risk. - Operators need preflight, go/no-go, monitoring, abort, cleanup, rollback, and verification steps. ## Do not use this skill when - The target data model or invariant is not yet designed; use `postgres-schema-design`. - The performance mechanism or exact index is unresolved; use `postgres-query-performance-review` or `postgres-index-design`. - The task is a broad review of current deployed objects; use `postgres-schema-review`. - The task is defining backup, PITR, retention, or recovery exercises; use `postgres-backup-recovery`. ## Inputs Inspect or obtain: - Exact PostgreSQL version/provider, extensions, topology, replicas/CDC, storage/WAL headroom, and operational limits. - Exact proposed DDL/data operations, dependenc