The Agent Skills Directory

Credential Handling in Documentation: (1) The reference files contain example commands and Infrastructure-as-Code templates with placeholder passwords, such as in SKILL.md and references/iac-usage.md. (2) While these are intended as examples, they are a security consideration if adopted without modification. The skill mitigates this by explicitly directing users toward Secret Manager and IAM database authentication for production environments.
External Tool References: (1) The documentation provides links to official vendor repositories and the mcp-toolbox.dev community project for development configuration. (2) Referencing external projects is a common practice for assisting users with environment setup, though it requires discretion when integrating third-party tools.
Command Execution Patterns: (1) The skill includes gcloud CLI commands for database administration tasks like cluster creation and scaling. (2) These patterns are intended for use by users or agents with appropriate credentials and represent standard administrative functionality within the Google Cloud ecosystem.
Interaction Surface for Indirect Prompt Injection: (1) The skill allows the agent to interpret user requests to perform database management tasks. (2) This creates a surface where external instructions could potentially influence agent actions. Evidence: * Ingestion points: User queries related to database tasks and MCP tool configurations defined in SKILL.md and references/mcp-usage.md. * Boundary markers: The Directives for Agents section in SKILL.md establishes clear constraints on how the agent should interpret and fulfill requests. * Capability inventory: The skill utilizes gcloud CLI commands for cluster, instance, and backup management as detailed in references/cli-usage.md. * Sanitization: The skill enforces strict directives that require specific connection methods (Auth Proxy) and identity roles (IAM), reducing the risk of unauthorized configuration attempts.

alloydb-basics