SQL Query Explainer

Explain a SQL query in plain English and flag performance and correctness risks.

What it does

Loads as a prompt for the InTouch AI assistant and turns it into a senior database engineer. Given a SQL statement, it restates what the query does, walks the logic in execution order (FROM/JOIN → WHERE → GROUP BY → HAVING → windows → SELECT → ORDER BY → LIMIT), and then audits the query for two classes of problems:

• Correctness risks — silent LEFT JOIN-to-inner-join filters, NULL comparison traps, NOT IN with nullable columns, aggregates mixed with non-grouped columns, DISTINCT hiding a fan-out duplication, integer division, non-deterministic ORDER BY ... LIMIT, and unscoped UPDATE/DELETE.
• Performance risks — non-sargable predicates (functions on indexed columns, leading-wildcard LIKE, implicit casts), SELECT *, correlated subqueries that should be joins, OR blocking index use, cartesian explosions, and unbounded result sets.

Each finding is rated [HIGH] / [MEDIUM] / [LOW], and HIGH/MEDIUM findings come with a concrete fix (rewrite, index, or predicate change).

What to feed it

• Required: the SQL query as plain text (any statement type — SELECT, INSERT, UPDATE, DELETE, MERGE, DDL).
• Optional: the SQL dialect (postgres, mysql, oracle, sqlserver, derby, …), the relevant table/index DDL, and the author's intended goal. Supplying these sharpens the analysis; without them the skill reasons from query structure and states its assumptions.

What it returns

A Markdown report with these sections: Summary, Plain-English Walkthrough, Correctness Risks, Performance Risks, an optional Suggested Rewrite, and Assumptions. When a category has no issues, it says so explicitly rather than inventing findings, and it never fabricates index names, row counts, or schema details.