The Agent Skills Directory

[Command Execution for Environment Setup]: The skill uses shell commands to authenticate with Google Cloud (gcloud auth login), enable APIs, and install dependencies. While these involve executing system-level commands, they are standard procedures for initializing the required cloud environment.
[Runtime Package Installation]: The environment setup and verify_all.sh script perform conditional installations of Python packages from scripts/requirements.txt if they are missing. This is a common pattern for ensuring required SDKs are available before executing inference samples.
[Model Availability Probing]: The skill includes a curl command to probe regional availability of Gemini models. This involves sending a minimal network request to the Google Cloud AI Platform API. This is used as a safety check to prevent users from attempting inference in regions where a model is not supported.
[Credential Management via SDKs]: The provided Python scripts demonstrate how to programmatically refresh OAuth access tokens using google-auth. This is a recommended practice for secure authentication to cloud services, replacing the need for hardcoded long-lived secrets.

agent-platform-inference