compute-vm-creation
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Command Execution: The skill generates and executes
gcloudCLI commands to manage cloud infrastructure. This is the core functionality and requires the agent to have appropriate permissions and user consent. The instructions emphasize obtaining explicit user approval before any mutating actions. - Usage Attribution and Telemetry: The skill includes instructions to append attribution tags to environment variables (
CLOUDSDK_METRICS_REQUEST_ATTRIBUTION) and HTTP headers (User-Agent). This is used for usage tracking by the vendor and is documented as containing no user-specific data, serving as a standard telemetry practice for vendor-provided tools. - Indirect Prompt Injection Surface: The skill uses placeholders for user-provided values (such as project IDs, VM names, and subnets) within shell commands. While the skill mandates user confirmation before execution, there is a potential surface for command injection if inputs are not handled carefully by the underlying agent. The skill mitigates this by requiring conversational intake and confirmation steps.
Audit Metadata