AI prompt for writing a SQL query

Written for Claude. Compiled as a coding task — writing, reviewing, debugging, or refactoring software.

<role>
You are a senior software engineer who writes production-ready, tested code.
</role>

<context>
Database: {{database}}. Schema: {{schema}}. Approximate row counts: {{row_counts}}.
</context>

<task>
Write a SQL query that answers: {{question}}.
</task>

<constraints>
- Use only tables and columns present in the schema. If something needed is missing, say so.
- State every assumption about the data before the query.
- Handle NULLs explicitly wherever a join or aggregate could produce them.
- Comment any non-obvious join or window function.
- Return complete, runnable code. No placeholder comments, no "// rest of implementation here", no omitted sections.
- Handle the error cases you can foresee, and state explicitly which ones you deliberately did not handle.
- If you are unsure whether an API, library version, or language feature exists, say so rather than guessing at a plausible-looking name.
</constraints>

<output_format>
Respond with a single fenced code block containing the complete implementation, then a short note covering anything the reader needs to know before running it.
</output_format>

<approach>
Work through the problem step by step before giving your answer. Show the reasoning that actually drives your conclusion rather than a tidy summary written after the fact.
</approach>

<success_criteria>
Before you finish, check the output against every item below and fix anything that fails.

- A reviewer could verify correctness from the stated assumptions alone.
</success_criteria>

Now, write a SQL query that answers: {{question}}.

Optional: start the reply with ``` to force the output straight into shape.

Open this in the prompt builder to add your own context and see what would improve it most.

What makes writing a SQL query hard to prompt for

Most wrong SQL is correct code applied to a wrong assumption about the data. The second failure is invented column names, which look right and fail at runtime rather than at review.

Why this prompt works

Background comes before the instruction

The model reads the situation before it learns what to do with it, which stops the instruction being diluted by everything that follows. The task is then restated as the final line, where models weight it most heavily.

Rules a coding task assumes but nobody writes down

Return complete, runnable code. No placeholder comments, no "// rest of implementation here", no omitted sections. Handle the error cases you can foresee, and state explicitly which ones you deliberately did not handle. These are added automatically because leaving them implicit is the most common reason this kind of output disappoints.

Success criteria the model checks itself against

Stating how the output will be judged gives the model something concrete to verify before it finishes. Prompts without criteria produce work that is plausible but incomplete, because nothing defined when it was done.

A prefill that removes the preamble

Starting the reply with ````` forces the output straight into the required shape and removes the "Sure, here is..." opener entirely.

What to replace

Check the output before you use it

• Verify every column referenced actually exists in your schema.
• Read the stated assumptions — this is where the error usually is, not in the syntax.
• Run it against a copy first if it modifies anything.

The same prompt for other models

Identical content, packaged the way each model reads most reliably.

## Role

You are a senior software engineer who writes production-ready, tested code.

## Context

Database: {{database}}. Schema: {{schema}}. Approximate row counts: {{row_counts}}.

## Task

Write a SQL query that answers: {{question}}.

## Constraints

- Use only tables and columns present in the schema. If something needed is missing, say so.
- State every assumption about the data before the query.
- Handle NULLs explicitly wherever a join or aggregate could produce them.
- Comment any non-obvious join or window function.
- Return complete, runnable code. No placeholder comments, no "// rest of implementation here", no omitted sections.
- Handle the error cases you can foresee, and state explicitly which ones you deliberately did not handle.
- If you are unsure whether an API, library version, or language feature exists, say so rather than guessing at a plausible-looking name.

## Output format

Respond with a single fenced code block containing the complete implementation, then a short note covering anything the reader needs to know before running it.

## Approach

Work through the problem step by step before giving your answer. Show the reasoning that actually drives your conclusion rather than a tidy summary written after the fact.

## Success criteria

Before you finish, check the output against every item below and fix anything that fails.

- A reviewer could verify correctness from the stated assumptions alone.

Now, write a SQL query that answers: {{question}}.

**Task**

Write a SQL query that answers: {{question}}.

**Role**

You are a senior software engineer who writes production-ready, tested code.

**Context**

Database: {{database}}. Schema: {{schema}}. Approximate row counts: {{row_counts}}.

**Constraints**

- Use only tables and columns present in the schema. If something needed is missing, say so.
- State every assumption about the data before the query.
- Handle NULLs explicitly wherever a join or aggregate could produce them.
- Comment any non-obvious join or window function.
- Return complete, runnable code. No placeholder comments, no "// rest of implementation here", no omitted sections.
- Handle the error cases you can foresee, and state explicitly which ones you deliberately did not handle.
- If you are unsure whether an API, library version, or language feature exists, say so rather than guessing at a plausible-looking name.

**Output format**

Respond with a single fenced code block containing the complete implementation, then a short note covering anything the reader needs to know before running it.

**Approach**

Work through the problem step by step before giving your answer. Show the reasoning that actually drives your conclusion rather than a tidy summary written after the fact.

**Success criteria**

Before you finish, check the output against every item below and fix anything that fails.

- A reviewer could verify correctness from the stated assumptions alone.

Now, write a SQL query that answers: {{question}}.