Database

Execute all three phases in order.

Phase 1 — Analysis

1. Take the slow query or ORM code.
2. Generate the equivalent raw SQL.
3. Either instruct the user to run EXPLAIN QUERY PLAN (SQLite) or EXPLAIN ANALYZE (Postgres/MySQL), or infer missing indexes directly from the WHERE, JOIN, and ORDER BY clauses.

Phase 2 — N+1 Audit

Check whether queries are issued inside a loop. If yes, rewrite using:

• IN (...) batch clause
• SQL JOIN
• ORM eager loading:
• .include() — Prisma
• .populate() — Mongoose
• select_related() / prefetch_related() — Django

Phase 3 — Rewrite

1. Provide the optimized SQL or ORM code.
2. Provide the exact CREATE INDEX statements required.
3. Explain the disk I/O reduction: Full Table Scan O(N) → Index Lookup O(log N).

Before/after query plan summary:

| Metric | Before | After | |---|---|---| | Scan type | Full Table Scan | Index Lookup | | Complexity | O(N) | O(log N) | | Query count (per page load) | N+1 | 1 |

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: v0idOS
• Source: v0idOS/performance-deity
• License: MIT
• Homepage: https://v0idos.github.io/performance-deity/

Install and usage instructions live in the source repository linked above.