progremir/pglr

Overview

This skill provides a secure PostgreSQL CLI tailored for AI agents to inspect and query databases without exposing connection credentials. It supports schema discovery, table listing, table description, and parameterized queries while enforcing read-only defaults and row limits. Use it to explore data structure and run safe queries during analysis tasks.

How this skill works

The CLI runs commands that return JSON-formatted results for queries, table listings, and schema inspection. It requires a human to configure the connection once; the skill cannot read or modify credentials. Queries default to read-only; write operations require an explicit flag from the user. Results include success status, row count, execution time, and optional truncation metadata.

When to use it

• Explore database structure before building queries or data pipelines
• List tables and inspect column definitions to design queries
• Run parameterized SELECT queries for analysis and reporting
• Safely query production-like databases without exposing credentials
• Check schema or table metadata during debugging or migrations

Best practices

• Always run schema or table description first (schema, describe <table>) to understand types and relationships
• Use parameterized queries to prevent SQL injection (pglr query with --params)
• Limit result size with LIMIT or rely on the 1000-row cap when sampling large tables
• Avoid write operations unless the user explicitly requests them and then use --allow-writes
• When multiple connections exist, specify --connection to target the correct database

Example use cases

• Quickly list all tables with pglr tables to find a target dataset
• Describe a table (pglr describe users) before composing a select for analytics
• Run a parameterized query to fetch recent orders for a dashboard
• Get a full schema overview (pglr schema) when onboarding to a new database
• Execute a controlled insert into logs only after explicit user approval with --allow-writes