The Agent Skills Directory

[Security Best Practices]: The skill proactively instructs the agent to enforce security measures, including the principle of least privilege for IAM roles, the use of the AlloyDB Auth Proxy for secure connections, and warnings against exposing databases to the public internet (0.0.0.0/0).
[External Resource Reference]: The skill references 'mcp-toolbox.dev', an external community resource for Model Context Protocol development. While this is a common tool for developers, it is an external site outside of the primary vendor's infrastructure.
[Indirect Prompt Injection Surface]: The skill processes user-supplied data (such as cluster IDs and instance names) to generate CLI commands and Terraform configurations. While no malicious behavior was detected, this pattern creates a surface for indirect prompt injection if inputs are not properly validated. * Ingestion points: User-provided variables in SKILL.md and references/cli-usage.md (e.g., CLUSTER_ID, REGION, PASSWORD). * Boundary markers: None explicitly defined in the templates. * Capability inventory: Execution of gcloud commands for resource management and database operations. * Sanitization: Not explicitly defined within the markdown documentation.

alloydb-basics