BigQuery Skills: BigQuery Expert

A comprehensive BigQuery plugin for Claude Code. Five integrated skill areas that activate automatically -- writing queries, designing schemas, optimizing costs, detecting anti-patterns, and navigating BigQuery-specific features.

Skills

Skill	Coverage
Query Generation	Generate optimized SQL from natural language. Convert queries from PostgreSQL, MySQL, Snowflake, Redshift, and SQL Server.
Query Optimization	Detect and fix 11 SQL anti-patterns with before/after rewrites. Project-wide scanning across `.sql` and code files.
Schema Design	Partitioning (time-unit, integer-range, ingestion-time), clustering, nested/repeated fields (STRUCT/ARRAY), denormalization, table types, and data type selection.
Cost Optimization	On-demand vs editions pricing, bytes-billed reduction, slot optimization, materialized views, query caching, storage management, and dry-run estimation.
BigQuery Features	STRUCT/ARRAY/UNNEST, MERGE DML, scripting, JSON functions, approximate aggregation, geography, BigQuery ML, search indexes, and vector search.

Skills activate based on context. Ask Claude to write a query and the generation skill engages. Discuss partitioning and the schema design skill kicks in. Multiple skills can activate simultaneously when a request spans areas.

Installation

Add the marketplace

/plugin marketplace add justvinhhere/bigquery-expert

Install the plugin

/plugin install bigquery-expert@justvinhhere-bigquery-expert

Or open /plugin, go to the Discover tab, and select bigquery-expert.

Activate

/reload-plugins

Commands

Command	What It Does
`/bigquery-expert:bq-generate`	Generate optimized SQL from a natural language description
`/bigquery-expert:bq-review`	Review SQL for performance anti-patterns
`/bigquery-expert:bq-optimize`	Rewrite SQL with all detected anti-patterns fixed
`/bigquery-expert:bq-design-table`	Design a table schema with partitioning, clustering, and data types
`/bigquery-expert:bq-estimate-cost`	Estimate the cost of a query or table
`/bigquery-expert:bq-explain`	Explain a BigQuery feature with working examples

All commands accept a file path, inline SQL, or a description as an argument. Without arguments, they use the most recent SQL in the conversation.

Examples:

/bigquery-expert:bq-generate "daily active users grouped by country for the last 30 days"
/bigquery-expert:bq-review path/to/query.sql
/bigquery-expert:bq-optimize "SELECT * FROM `project.dataset.events`"
/bigquery-expert:bq-design-table "user click events with timestamp, page URL, and session ID"
/bigquery-expert:bq-estimate-cost path/to/expensive_query.sql
/bigquery-expert:bq-explain "MERGE for upserts"

Agents

Agents run autonomously across your project when you ask naturally:

Agent	Use When You Say...
bq-reviewer	"Review all SQL files in this project for anti-patterns"
bq-schema-advisor	"Audit my table schemas and recommend partitioning strategies"
bq-cost-analyzer	"Which queries in this project are the most expensive?"

Anti-Pattern Detection

The query optimization skill detects 11 BigQuery SQL anti-patterns:

#	Pattern	Fix	Severity
1	`SELECT *` on single-table query	Specify only needed columns	High
2	`IN`/`NOT IN` without `DISTINCT`	Add `DISTINCT` to subquery	Medium
3	CTE referenced multiple times	Convert to `CREATE TEMP TABLE`	High
4	`ORDER BY` without `LIMIT`	Add `LIMIT` clause	Medium
5	`REGEXP_CONTAINS` for simple patterns	Use `LIKE` instead	Low
6	`ROW_NUMBER()` + `WHERE rn = 1`	Use `ARRAY_AGG(... LIMIT 1)`	High
7	Subquery inside WHERE	Extract to `DECLARE` variable or CTE	Medium
8	WHERE predicates not ordered by selectivity	Reorder by operator cost (advisory)	Low
9	Smaller table first in JOIN	Place largest table first (advisory)	Low
10	`CREATE TEMP TABLE` without `DROP`	Add `DROP TABLE` at end of script	Low
11	`CREATE TABLE` + `DROP TABLE` in same script	Use `CREATE TEMP TABLE` instead	Low

Based on BigQuery Anti-Pattern Recognition by Google Cloud Platform (Apache 2.0).

Example: Before and After

Before -- ROW_NUMBER() for latest record per group (High severity):

SELECT taxi_id, trip_seconds, fare
FROM (
  SELECT taxi_id, trip_seconds, fare,
    ROW_NUMBER() OVER (PARTITION BY taxi_id ORDER BY fare DESC) rn
  FROM `bigquery-public-data.chicago_taxi_trips.taxi_trips`
)
WHERE rn = 1

After -- ARRAY_AGG (optimized):

SELECT event.*
FROM (
  SELECT ARRAY_AGG(
    t ORDER BY t.fare DESC LIMIT 1
  )[OFFSET(0)] event
  FROM `bigquery-public-data.chicago_taxi_trips.taxi_trips` t
  GROUP BY t.taxi_id
)

Plugin Contents

skills/
  bigquery-optimization/       11 anti-pattern references
  bigquery-query-generation/   Schema-aware generation, common patterns, dialect conversion
  bigquery-schema-design/      Partitioning, clustering, nested fields, denormalization, types
  bigquery-cost-optimization/  Pricing, bytes-billed, slots, materialized views, storage
  bigquery-features/           STRUCT/ARRAY, MERGE, scripting, JSON, geo, BQML, vector search
commands/
  bq-generate, bq-review, bq-optimize, bq-design-table, bq-estimate-cost, bq-explain
agents/
  bq-reviewer, bq-schema-advisor, bq-cost-analyzer

Uninstall

/plugin uninstall bigquery-expert@justvinhhere-bigquery-expert
/plugin marketplace remove justvinhhere-bigquery-expert

Contributing

Contributions welcome. Fork the repository, create a feature branch, and submit a pull request.

For bugs and feature requests, open an issue.

License

Apache License 2.0 -- see LICENSE for details.