The Agent Skills Directory

Command Execution: The skill uses Google Cloud CLI (gcloud) commands to discover environment details, such as the active project ID and existing cloud resources like Storage buckets and GKE clusters. These operations are local to the user's authenticated environment and are necessary for the skill's resource discovery functionality.
Dynamic Execution: Based on user requirements, the skill generates bash scripts and Terraform HCL configurations. It provides instructions to execute these scripts to deploy infrastructure. This dynamic generation and execution are the primary functions of the skill and utilize official vendor-supported deployment methods.
Indirect Prompt Injection Surface: The skill processes user-provided inputs, such as architecture names and routing rules, and interpolates them into deployment scripts and configurations.
Ingestion points: User-defined names, descriptions, and routing rules, as well as the output of resource discovery commands.
Boundary markers: The instructions do not explicitly specify delimiters for the interpolated data within the generated code.
Capability inventory: The skill involves writing configuration files to the workspace and executing shell commands via the Google Cloud SDK.
Sanitization: The skill implements validation logic, such as converting resource names to lowercase and enforcing regex patterns, to mitigate potential injection risks in the generated infrastructure code.
File System Operations: The skill includes instructions to manage local directories, such as clearing local Terraform state folders to ensure deployment consistency. These are standard maintenance tasks for infrastructure-as-code workflows.

google-cloud-global-frontend-configuration