Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

SQL Query Generator · phuryn | DecimalAI

phuryn/sql-queries

SkillScore91Maintained

Generate SQL queries from natural language descriptions. Supports BigQuery, PostgreSQL, MySQL, and other dialects. Reads database schemas from uploaded diagrams or documentation. Use when writing SQL, building data reports, exploring databases, or translating business questions into queries.

Databy phuryn

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Data Analyst pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/phuryn-sql-queries/SKILL.md

decimalai skills pull phuryn-sql-queries --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/phuryn-sql-queries/SKILL.md https://app.decimal.ai/s/phuryn-sql-queries/SKILL.md

Raw file: /s/phuryn-sql-queries/SKILL.md ↗ · machine-readable summary at /s/phuryn-sql-queries.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 54d ago

Yes — +5% lift · at higher cost+105% tokens0% turns91% of whole cases pass with skill (20/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-20	✗→✓	▲ Improved	68%	0%
case-02	✓→✓	＝ Same ✓	102%	0%
case-03	✓→✓	＝ Same ✓	116%	0%
case-04	✓→✓	＝ Same ✓	150%	0%
case-05	✓→✓	＝ Same ✓	145%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

phuryn / pm-skills

pm-data-analytics/skills/sql-queries/SKILL.md

View on GitHub ↗

License MITImported 8/3/2026Version v1

Skill contents

SKILL.md

SQL Query Generator

Purpose

Transform natural language requirements into optimized SQL queries across multiple database platforms. This skill helps product managers, analysts, and engineers generate accurate queries without manual syntax work.

How It Works

Step 1: Understand Your Database Schema

If you provide a schema file (SQL, documentation, or diagram description), I will read and analyze it
Extract table names, column definitions, data types, and relationships
Identify primary keys, foreign keys, and indexing strategies

Step 2: Process Your Request

Clarify the exact data you need to retrieve or analyze
Confirm the SQL dialect (BigQuery, PostgreSQL, MySQL, Snowflake, etc.)
Ask for any additional requirements (filters, aggregations, sorting)

Step 3: Generate Optimized Query

Write efficient SQL that leverages your database structure
Include comments explaining complex logic
Add performance considerations for large datasets
Provide alternative approaches if applicable

Step 4: Explain and Test

Explain the query logic in plain English
Suggest how to test or validate results
Offer tips for performance optimization
If you want, generate a test script or sample data

Usage Examples

Example 1: Query from Schema File

Upload your database_schema.sql file and say:
"Generate a query to find users who signed up in the last 30 days
and had at least 5 active sessions"

Example 2: Query from Diagram Description

"Here's my database: Users table (id, email, created_at), Sessions table
(id, user_id, timestamp, duration). Generate a query for average session
duration per user in January 2026."

Example 3: Complex Analysis Query

"Create a BigQuery query to analyze our revenue by region and customer tier,
including year-over-year growth rates."

Key Capabilities

Multi-Dialect Support: Works with BigQuery, PostgreSQL, MySQL, Snowflake, SQL Server
File Reading: Reads schema files, SQL dumps, and data documentation
Query Optimization: Suggests indexes, partitioning, and performance improvements
Explanation: Breaks down queries for learning and documentation
Testing: Can generate test queries and sample data scripts
Script Execution: Create executable SQL scripts for your database

Tips for Best Results

Provide context: Share your database schema or structure
Be specific: Clearly describe what data you need and any filters
Mention database: Specify which SQL dialect you're using
Include constraints: Mention data volume, time ranges, and performance needs
Request format: Ask for the query result format if you need specific output

Output Format

You'll receive:

SQL Query: Production-ready SQL code with comments
Explanation: What the query does and how it works
Performance Notes: Optimization tips and considerations
Test Script (if requested): Sample data and validation queries

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 Write a PostgreSQL query for a user retention report on the `users` and `sessions` tables. Filter for accounts created in the last 30 days relative to CURRENT_DATE without wrapping `users.created_at` in a function, and return only users who have logged at least 5 distinct sessions.	fail→fail	7,375	7,545	+2%	1	1	0%	1,284	2,300	+79%	0	0	—
case-02 Construct a Google BigQuery SQL query calculating year-over-year revenue percentage growth per customer from annual summary data in table `customer_sales` (columns: `customer_id`, `year`, `revenue`). Use `LAG` for prior year revenue and handle zero revenue safely without division-by-zero errors.	pass→pass	6,825	9,930	+45%	1	1	0%	1,409	2,844	+102%	0	0	—
case-03 Write a Snowflake SQL query calculating the average session duration per user for all sessions recorded during January 2026. The table `sessions` has `user_id`, `session_timestamp` (TIMESTAMP_NTZ), and `duration` (seconds). Ensure the date range filter includes all session times on January 31st without using functions on `session_timestamp`.	pass→pass	6,498	8,037	+24%	1	1	0%	1,130	2,441	+116%	0	0	—
case-04 Write a MySQL query to group registered users in the `users` table into monthly cohorts using `created_at`. Format the cohort column as a first-of-the-month date string formatted precisely as 'YYYY-MM-01' using MySQL date formatting.	pass→pass	3,599	6,698	+86%	1	1	0%	861	2,155	+150%	0	0	—
case-05 Write a T-SQL query for Microsoft SQL Server retrieving the single most recent order per customer from `orders` (columns: `customer_id`, `order_date`, `order_id`). Use a Common Table Expression named `ranked_orders` with `ROW_NUMBER()` partitioned by `customer_id` and ordered by `order_date` descending, breaking ties deterministically with `order_id` descending.	pass→pass	2,994	4,679	+56%	1	1	0%	688	1,685	+145%	0	0	—
case-06 Write a PostgreSQL query to retrieve `customer_id` from the `events` table where the JSONB column `metadata` has a top-level text field `'tier'` equal to `'enterprise'`. Use the text extraction operator rather than JSON containment or JSON path operators.	pass→pass	5,024	2,947	-41%	1	1	0%	519	1,310	+152%	0	0	—
case-07 Write a BigQuery SQL query to count records in `events`, which is partitioned by the `event_date` DATE column, for the exact month of February 2026. Filter `event_date` directly using literal DATE strings without using `BETWEEN` or functions on `event_date`.	fail→fail	7,032	4,878	-31%	1	1	0%	1,215	1,958	+61%	0	0	—
case-08 Write a Snowflake SQL query to retrieve all columns from the `logs` table, keeping only the most recent record for each `tracking_id` based on `created_at` DESC. Use Snowflake's `QUALIFY` clause with `ROW_NUMBER()` directly in the main statement without writing a subquery or CTE.	pass→pass	3,529	4,120	+17%	1	1	0%	608	1,599	+163%	0	0	—
case-09 Write a MySQL query filtering the `orders` table for `status = 'completed'` during 2025. Explicitly enforce index usage using the `FORCE INDEX` directive for the index named `idx_order_status_date`.	pass→pass	5,030	6,291	+25%	1	1	0%	985	2,015	+105%	0	0	—
case-10 Write a PostgreSQL query to build an organizational tree from `employees` (columns: `employee_id`, `manager_id`, `name`). Define a recursive Common Table Expression named `org_tree` starting with top executives where `manager_id IS NULL`, computing an integer column `level` starting at 1 for executives and incrementing by 1 for each child level.	pass→pass	4,400	10,267	+133%	1	1	0%	960	2,409	+151%	0	0	—
case-11 Write a Google BigQuery SQL query to calculate an approximate distinct count of `visitor_id` from the `traffic_logs` table for fast hyperloglog estimation on massive datasets.	pass→pass	5,787	9,370	+62%	1	1	0%	1,097	2,464	+125%	0	0	—
case-12 Write a Snowflake SQL query selecting all rows from `accounts` as the table existed exactly 2 hours (7200 seconds) ago, using Snowflake Time Travel with a negative second offset.	pass→pass	3,671	5,430	+48%	1	1	0%	641	1,425	+122%	0	0	—
case-13 Write a MySQL query grouping `products` by `category_id`, generating a single string of product names per category named `product_list`. Format the list as comma-space separated items (`', '`) ordered alphabetically by `product_name` within each group using MySQL's `GROUP_CONCAT`.	pass→pass	4,727	4,202	-11%	1	1	0%	682	1,622	+138%	0	0	—
case-14 Write a T-SQL query for Microsoft SQL Server joining `customers c` with a correlated subquery on `orders o` using `CROSS APPLY` to fetch each customer's top 2 highest total_amount orders.	pass→pass	4,628	5,741	+24%	1	1	0%	885	1,878	+112%	0	0	—
case-15 Write a PostgreSQL query joining `categories c` to `products p` using a `LATERAL` subquery to select the 3 cheapest products (`price ASC`) for each category.	pass→pass	5,016	6,664	+33%	1	1	0%	936	1,986	+112%	0	0	—
case-16 Write a BigQuery SQL query to flatten the `tags` ARRAY column in the `articles` table and return each individual `tag` alongside its total frequency count across all articles, ordered by count descending.	pass→pass	3,360	6,286	+87%	1	1	0%	654	1,869	+186%	0	0	—
case-17 Write a Snowflake SQL query to unpack elements from an ARRAY stored inside a VARIANT column `items_purchased` in the `orders` table using Snowflake's table function for array flattening.	pass→pass	4,717	9,666	+105%	1	1	0%	934	2,136	+129%	0	0	—
case-18 Write a MySQL statement to insert a record into `user_stats` (columns: `user_id`, `login_count`) with values (123, 1). If a key collision occurs on `user_id`, increment the existing `login_count` by 1 using MySQL's upsert clause.	pass→pass	5,052	4,361	-14%	1	1	0%	826	1,641	+99%	0	0	—
case-19 Write a PostgreSQL 14+ query that aggregates web request counts from `web_logs` into 15-minute time bins using PostgreSQL 14's native `DATE_BIN` function. Align the bins to the origin timestamp `TIMESTAMP '2026-01-01 00:00:00'` on the `request_time` column.	pass→pass	3,776	8,202	+117%	1	1	0%	868	2,608	+200%	0	0	—
case-20 Provide the Liquibase XML changeSet syntax to add a foreign key constraint named fk_order_user between orders.user_id and users.id in PostgreSQL.	fail→pass	5,255	6,898	+31%	1	1	0%	1,106	1,854	+68%	0	0	—
case-21 Write a Prisma schema definition for a 1-to-many relationship between User and Post models.	pass→pass	4,451	6,302	+42%	1	1	0%	883	1,977	+124%	0	0	—
case-22 What postgresql.conf settings should be configured on a dedicated PostgreSQL server with 64GB RAM to improve general query performance?	pass→pass	15,445	16,414	+6%	1	1	0%	3,043	3,406	+12%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 86.4%
Pass rate with the skill: 90.9%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 8/4/2026, 19:05 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-20

Prompt

Provide the Liquibase XML changeSet syntax to add a foreign key constraint named fk_order_user between orders.user_id and users.id in PostgreSQL.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run91

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+112% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 54 days ago

Other measured skills in the registry, with their headline benchmark lift.