The Agent Skills Directory

Command Execution Framework: The skill instructs the agent to execute various gcloud commands for infrastructure management. To mitigate risks associated with direct command execution, the skill implements a mandatory safety tier system. High-risk operations, such as modifying IAM policies (Tier B) or deleting log data (Tier D), require explicit interactive or typed confirmation from the user before execution.
User-Controlled Data Ingestion: The skill uses placeholders (e.g., {project_id}, {filter_expression}) to incorporate user input into shell commands. This represents a potential surface for indirect prompt injection or command injection. However, the skill's design addresses this by requiring manual review and confirmation for sensitive commands, ensuring the user can inspect the final command string before it is run.
External Documentation References: The skill references several official Google Cloud documentation pages. These are used to provide the user and the agent with legitimate technical context and installation instructions for the required tools.

cloud-logging-configuration-basics