Audit my database for unused and redundant indexes and remove the ones that are only slowing down writes. 1. Find unused indexes: query the database's index usage statistics (pg_stat_user_indexes for Postgres, the equivalent for MySQL or my ORM's introspection tools) and list every index with zero or near-zero scans since the stats were last reset. 2. Find redundant indexes: identify indexes that are fully covered by another index (an index on (a) is redundant if one on (a, b) exists, since the composite can serve queries on a alone) and duplicate indexes created twice under different names. 3. Verify before dropping: for each candidate, check my actual query patterns first. Confirm no query in the codebase filters or joins on those columns in a way that would lose its index, and rule out indexes that only serve rare-but-important queries (nightly jobs, admin reports) before removing them. Also confirm the index is not enforcing a constraint (unique indexes stay). 4. Drop safely: write the migration to drop the confirmed-dead indexes concurrently where the database supports it, so the drop itself does not lock the table. Do not drop unique indexes or indexes backing foreign keys without flagging them separately. Do not trust stats alone if the database was restarted recently; tell me how old the stats are. Give me a table: index name, size, scan count, verdict (drop, keep, investigate), and why. Then the migration for the drops.