query-validation

name	query-validation
description	Review SQL queries for correctness, performance, and best practices before they reach production. Activate when promoting queries to dashboards, investigating unexpected results, or when a query is slow and needs optimisation.

When to use

• Promoting a query to a production dashboard or scheduled report
• Results look wrong or unexpected — validating logic before escalating
• A query is slow and needs performance investigation
• Preventing anti-patterns from causing production incidents
• Code review for an analyst's or engineer's SQL

Process

1. Syntax & style check — run scripts/sql_lint.py (uses sqlglot) to catch syntax errors, unsupported functions for the target DB, and style violations
2. Anti-pattern detection — compare against references/sql_anti_patterns.md; rate each finding by severity (blocking / major / minor)
3. Execution plan analysis — parse EXPLAIN or query profile output via scripts/explain_plan_parser.py; identify full table scans, missing indexes, and sort spills
4. Cardinality estimation — run scripts/cardinality_estimator.py when schema statistics exist; flag joins with unexpected row multiplication
5. Engine-specific validation — cross-reference references/engine_specific_guide.md for Snowflake, BigQuery, Postgres, or Redshift-specific behaviour (date functions, window frame defaults, QUALIFY support)
6. Produce deliverables — populate assets/query_review_template.md with categorised findings; generate assets/optimization_recommendations.md for performance issues

Inputs the skill needs

• Required: SQL text
• Required: target database engine (Snowflake / BigQuery / Postgres / Redshift / other)
• Should provide: table schemas and approximate row counts
• Should provide: EXPLAIN or query profile output if the query has already run
• Optional: business logic the query is meant to implement
• Optional: performance target (e.g. must run under 30 s on prod)

Output

• assets/query_review_template.md (filled) — correctness, performance, style findings categorised by severity
• assets/optimization_recommendations.md (filled) — prioritised optimisation suggestions with projected impact