The Agent Skills Directory

Command Execution: The skill instructs the agent to execute shell commands using agents-cli and gcloud. These commands are used to manage agent registrations, list resources, and update service metadata within Google Cloud. While these are intended functionalities for the tool, they involve interaction with the underlying system.
External Package Installation: The skill references the installation of google-agents-cli using the uv tool. This package appears to be a resource provided by the vendor (Google) for managing agent deployments. Following standard security practices, users should ensure they are using official packages from trusted registries.
Indirect Prompt Injection Surface: The skill includes mechanisms to ingest data from local files like deployment_metadata.json and uv.lock, as well as fetching content from external URLs provided via the --agent-card-url flag.
Ingestion points: deployment_metadata.json, uv.lock, and remote agent card JSON files.
Boundary markers: The skill instructions do not explicitly define delimiters or specific sanitization steps for the data processed from these external sources.
Capability inventory: The tool has the capability to execute shell commands via agents-cli and gcloud based on the configuration data it processes.
Sanitization: There is no specific mention of input validation or sanitization for the external agent cards or metadata files within the skill instructions.

google-agents-cli-publish