database-index-designlisted

# Database Index Design ## Purpose Choose an index set that supports the measured workload at justified write, storage, locking, maintenance, and rollout cost. Retain an adequate portfolio; fewer indexes can still cost more if consolidation widens keys or loses useful access paths. This skill owns the portfolio and DDL decision; `sql-query-performance` owns reading one statement's executed plan. ## Required inputs ```text engine, exact version/edition, table shape, row count, and growth: representative query workload with frequency and tail parameter distributions: equality, range, join, ordering, projection, and uniqueness requirements per query: current indexes, constraints, usage window, write/update rate, and replica roles: executed plans and actual work for the target statements: DDL availability, lock, log/WAL, disk, rollback, and maintenance constraints: ``` Inspect existing schema/migrations, captured plans and usage windows before asking for new evidence. If the workload or engine is unknown, do not emit DDL; identify the missing decision input. Estimated plans can support candidates when actuals are unavailable, with runtime benefit explicitly unverified. This skill has no Java language minimum: its compatibility boundary is the database and migration tooling. In Java projects inspect resolved JDBC/ORM versions, generated SQL and parameter types, plus migration transaction settings. The references use PostgreSQL 18 and MySQL 8.4 for version-sensitive examples;