compute-on-demand-reservations
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFE
Full Analysis
- Trusted Official Tooling: The skill exclusively utilizes the
gcloudCLI and Google Cloud APIs to manage infrastructure. These are official, well-known services from the vendor, and the skill correctly directs users to official documentation for installation. - Human-Approval Guardrails: A strong security posture is maintained by explicitly instructing the agent to obtain human approval before executing any mutating commands (creation, update, or deletion of resources) or commands that incur billing charges.
- Usage Attribution and Metrics: The skill implements a tracking mechanism via the
CLOUDSDK_METRICS_REQUEST_ATTRIBUTIONenvironment variable and customUser-Agentheaders. These are documented as carry no user data and are used for service attribution within the vendor's ecosystem. - IAM and Least Privilege Guidance: The instructions provide detailed guidance on the specific IAM roles required for different operations, encouraging the use of least-privilege principles and custom service accounts for VM provisioning.
- Network Security Best Practices: For instance creation, the skill recommends using private networking options (such as
--no-addressand Identity-Aware Proxy) to minimize exposure. - Input Validation and Schema Awareness: The skill includes specific logic to handle variations in MCP (Model Context Protocol) tool schemas, ensuring that the agent falls back to manual CLI commands when tool parameters are missing or inconsistent.
Audit Metadata