The Agent Skills Directory

Command Execution and Telemetry: The skill instructs the agent to use gcloud CLI commands and REST API calls for database management. It includes a usage attribution mechanism that adds a skill-specific identifier to environment variables (CLOUDSDK_METRICS_REQUEST_ATTRIBUTION) and User-Agent headers. This is a standard telemetry practice for technical support and performance tracking, is transparently documented in references/attribution.md, and includes a user opt-out instruction.
External Package Dependencies: The skill references several official Google Cloud client libraries for Python, Java, Node.js, and Go. These are standard, well-maintained packages from a trusted vendor used for their intended purpose of database interaction.
Destructive Operation Safety: The skill includes a safety guardrail in SKILL.md that explicitly requires obtaining user confirmation before executing any non-emulator database changes (DML or DDL) or destructive operations. This helps prevent accidental data loss or unauthorized schema changes.
Indirect Prompt Injection Surface Area: As a database management tool, the skill naturally ingests data from external sources such as database schemas and query results (e.g., SPANNER_SYS tables). While this presents a theoretical surface for indirect prompt injection if the database content is attacker-controlled, the skill mitigates this by requiring human confirmation for sensitive actions and providing clear documentation on IAM security and the principle of least privilege.

spanner-basics