The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the installation of client libraries and tools from various external sources.
Instructions for installing cloud-sql-python-connector via pip in references/client-library-usage.md.
Instructions for installing @google-cloud/cloud-sql-connector via npm in references/client-library-usage.md.
Links to external GitHub repositories for Cloud SQL extensions under the gemini-cli-extensions organization in references/mcp-usage.md.
Links to mcp-toolbox.dev for integration sources in references/mcp-usage.md.
[COMMAND_EXECUTION]: The skill provides numerous examples of shell commands for managing cloud infrastructure.
Extensive list of gcloud sql commands for creating, describing, and modifying database instances in SKILL.md and references/cli-usage.md.
Instructions for running the cloud-sql-proxy binary to establish secure connections in SKILL.md.
[PROMPT_INJECTION]: The skill describes an architecture that creates an indirect prompt injection surface by ingesting data from managed databases.
Ingestion points: The execute_sql tool (referenced in references/mcp-usage.md) allows an agent to retrieve data from a Cloud SQL instance.
Boundary markers: There are no instructions or system prompts provided to the agent to treat the output of SQL queries as untrusted data or to ignore instructions contained within that data.
Capability inventory: The skill provides access to administrative commands (gcloud sql) and data manipulation tools (execute_sql) across its reference files.
Sanitization: The documentation does not include guidance on sanitizing or validating database output before it is processed by the AI agent.

cloud-sql-basics