The Agent Skills Directory

Command Execution: The skill uses subprocess.run() in scripts/lib/monitoring.py to invoke the gcloud CLI tool. This is a common pattern for interacting with cloud provider services and is used here to obtain authentication tokens and fetch environment configurations.
External Resource Access: The skill fetches telemetry data from the official Google Cloud Monitoring API (monitoring.googleapis.com) and provides links to official Google Cloud documentation for SDK installation and access control. These are considered trusted sources within the context of managing Google Cloud environments.
Data Ingestion Surface (Indirect Prompt Injection): The skill processes environment logs and configuration files using gcloud logging read and gcloud composer environments describe. While these sources could theoretically contain untrusted content, the skill includes instructions for structured analysis and evidence-based recommendations, and the read-only nature of the diagnostic scripts reduces the potential impact of such findings.
Authentication & Permissions: The skill relies on Application Default Credentials and standard IAM roles (roles/monitoring.viewer, roles/logging.viewer) to access telemetry. It explicitly includes constraints to never apply changes without user approval, adhering to the principle of least privilege for diagnostic tasks.

managed-airflow-recommendations