The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the use of the gcloud CLI to manage cloud resources. To mitigate the risk of incorrect or harmful execution, the instructions mandate the use of gcloud help <command> for syntax validation before any command is proposed or executed. It also enforces the use of the --quiet flag to ensure non-interactive, predictable behavior and requires explicit project and location flags to prevent accidental execution against incorrect environments.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a surface where data from gcloud command outputs (help text and resource listings) is ingested by the agent to inform subsequent actions.
Ingestion points: External data enters the agent context through the outputs of gcloud help and gcloud <group> <resource> list (as seen in the Discovery Workflow).
Capability inventory: The agent has the capability to execute shell commands via the gcloud binary.
Boundary markers: The skill defines a "Mandatory Plan Template" and a "Prohibited Operations" denylist that require human authorization for sensitive tasks, creating a clear boundary between automated discovery and resource modification.
Sanitization: The skill explicitly forbids the use of web search for CLI documentation, relying solely on local, official CLI help output to prevent the ingestion of potentially manipulated external instructions.
[EXTERNAL_DOWNLOADS]: The skill provides links to official installation guides and authorization documentation on docs.cloud.google.com. These references are directed to the trusted vendor's official documentation infrastructure for the purpose of tool setup and security configuration.

gcloud