Data
Audit your indexes
Which are unused, which are duplicates, and which are missing.
Fill it in
Where to look.
They cost every write and nobody removes them.
From the query statistics, not from guessing.
Your prompt
Audit the indexes on [the database]. Find the unused ones from the statistics. Every index costs on write and in storage, and nobody removes them because removing an index feels risky while adding one does not. Report scans since reset alongside size, so the cost is visible. Find the duplicates and the redundant prefixes: an index on (a) is redundant when one on (a, b) exists, and both are usually there because two people added them. Suggest missing indexes from the query statistics, not from reading the schema. A suggestion with no query behind it is a guess that costs writes forever. For each recommendation say what it costs and what it saves, and flag the ones that need a concurrent build.
Use Audit your indexesOpens with everything above already filled in.
Why this works
Nobody removes an index, because removing one feels risky while adding one does not, so every mature database carries writes it does not need. This reports scans against size, finds redundant prefixes, and only suggests new indexes with a query behind them.