The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the gcloud CLI to perform discovery and management of Google Cloud resources, including gcloud compute, gcloud storage, and gcloud infra-manager. It performs file system operations such as rm -rf .terraform to clear local states and instructs the agent to help users execute generated bash scripts (./deploy.sh) to actuate deployments.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external Google Cloud environments, creating an attack surface for indirect prompt injection.
Ingestion points: Data enters the context via references/resource-discovery.md, which fetches lists of existing Storage buckets, Managed Instance Groups, and Cloud Run services.
Boundary markers: The instructions do not define explicit delimiters to isolate external resource names from the agent's logic.
Capability inventory: The environment allows significant capabilities including shell command execution (gcloud), script creation, and file system modification.
Sanitization: The skill implements naming constraints using a regex (^[a-z](?:[-a-z0-9]{0,61}[a-z0-9])?$) for Terraform resource generation, which helps mitigate basic injection attempts.
[EXTERNAL_DOWNLOADS]: The skill fetches infrastructure-as-code configurations from official providers.
Evidence: It references official GoogleCloudPlatform modules for Load Balancing and Cloud Armor via the Terraform Registry and provides links to official documentation at cloud.google.com.

google-cloud-global-frontend-configuration