text-to-sql

The Text-to-SQL skill is a specialized agentic capability designed to bridge the gap between human-readable natural language and relational database operations. It serves developers, data analysts, and system administrators by translating intent into valid, performant, and secure SQL syntax. By integrating deeply with database-architect schemas, the skill understands table relationships, foreign keys, and indexing strategies to ensure queries are not only syntactically correct but also execution-optimized for the target environment.

• Advanced Query Generation: Transforms complex natural language prompts into precise SQL queries, including support for joins, aggregations, subqueries, and conditional filtering.

• Schema-Driven Accuracy: Directly references database-architect definitions to ensure table and column validation, minimizing hallucinations and runtime errors.

• Security-First Design: Mandatory use of parameterized queries to prevent SQL injection, alongside syntax validation and test-run capabilities.

• Performance Optimization: Automatically considers join efficiency, index usage, and data transfer constraints to generate performant code.

• Evaluation Integration: Compatible with promptfoo testing frameworks, allowing for rigorous assessment of syntax accuracy, functional correctness, and schema compliance.

• Provide a comprehensive database schema including table definitions, column types, and constraints to ensure high-fidelity outputs.

• Utilize chain-of-thought prompting for complex multi-table joins or analytical reporting queries to ensure logic flow is sound.

• Always set explicit limits on SELECT operations to protect database stability during ad-hoc exploration.

• Treat the output as a parameterized query; avoid manual string concatenation to maintain compliance with safety protocols.

• Best suited for tasks involving reporting, data exploration, audit logging, and ad-hoc troubleshooting within production or development database environments.