The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill ingests and processes data from external Google Cloud resource metadata and error messages, specifically within scripts like diagnose_error.py and manage_workload_updates.py. If an environment contains resources with malicious instructions in their names or descriptions, there is a potential risk of indirect prompt injection.
Ingestion points: API responses and gcloud command outputs are processed by scripts in the scripts/ directory.
Boundary markers: The skill relies on structured JSON and regular expressions for parsing, though it does not implement specific delimiters for the raw data ingested into the agent context.
Capability inventory: The skill can perform mutating operations including creating and deleting folders, projects, and workloads.
Sanitization: The validate_params.py script employs shlex.quote to sanitize parameters before they are used in shell commands, which is a significant mitigation against command injection.
[Administrative Cloud Operations]: The skill is designed to manage sensitive Google Cloud infrastructure, including folder and project deletion and organization policy updates. It requires high-level IAM roles such as roles/assuredworkloads.admin. While these are expected for the skill's purpose, the ability to perform destructive operations on cloud resources represents an administrative capability surface that should be monitored.
[External Dependency Usage]: The skill utilizes official Google Cloud Python libraries (google-cloud-assured-workloads, google-api-python-client, and google-auth). These are trusted resources from a well-known vendor and are appropriate for interacting with the Google Cloud API.
[Usage Telemetry]: The skill includes usage attribution in its gcloud and API requests through the CLOUDSDK_METRICS_REQUEST_ATTRIBUTION environment variable and custom User-Agent headers. This is a standard telemetry practice for identifying usage patterns related to specific tools.

assured-workloads-basics