Unlock Faster, Smarter Cloud SQL Management with Gemini CLI’s Agent Skills

Agent Skills Overview

Agent Skills act as a lightweight instruction set for Gemini. Instead of loading every possible command at startup, Gemini loads only the names and short descriptions of available skills. When a user request matches a description, the full SKILL.md file is loaded and executed.

Progressive Loading Mechanism

Discovery : On startup the Agent loads only skill names and brief descriptions.

Activation : When a request matches a description, the Agent loads the detailed instruction file into context.

Execution : The Agent runs the associated script, loading any referenced files or code as needed.

Refactored Cloud SQL for PostgreSQL Extension

The extension now uses a fine‑grained skills/ directory instead of a monolithic toolset. Each skill is a self‑contained package with its own script.

cloud-sql-postgres-admin : Manage tasks such as configuring new Cloud SQL instances.

cloud-sql-postgres-lifecycle : Handle the lifecycle of Cloud SQL instances.

cloud-sql-postgres-data : Explore database schemas.

cloud-sql-postgres-monitor : Diagnose performance issues.

cloud-sql-postgres-health : Audit database health.

cloud-sql-postgres-view-config : Discover and manage extension or engine‑level settings.

cloud-sql-postgres-replication : Monitor replication health.

Standardized Execution

Each skill maps to a script inside its package, ensuring the same execution logic works on Linux, macOS, and Windows.

Comparison with Traditional Tools

Loading mode : Traditional tools are eagerly loaded (all definitions at startup); Agent Skills are loaded progressively, only when needed.

Context usage : Eager loading consumes many tokens, risking context saturation; progressive loading keeps the token footprint minimal, preserving model reasoning accuracy.

Installation and Configuration

Built‑in installation : Running gemini extensions install places skills automatically under ~/.gemini/extensions/<extension_name>/skills.

Environment consistency : Existing environment variables are passed to skills at runtime.

Easy updates : Use gemini extensions config <extension_name> to modify extension variables.

Note: Pre‑loaded MCP servers in these extensions are deprecated in favor of the leaner skill‑based discovery mechanism.

Listing Available Skills

/skills list

Using Skills

List all tables in my production database.

Which tables in my Postgres instance consume the most space?

Run this SQL and give me a summary.

Gemini matches the appropriate skill, runs the underlying script, and returns the result.

Getting Started

Update Gemini CLI to v0.29.0+: npm install -g @google/gemini-cli@latest Update the Cloud SQL for PostgreSQL extension to v0.2.3+: gemini extensions update cloud-sql-postgresql If the extension is not installed, run:

gemini extensions install https://github.com/gemini-cli-extensions/cloud-sql-postgresql

Future Plans

The progressive skill model will be evaluated for other database extensions in the Gemini ecosystem.