The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's Python scripts (e.g., scripts/cloud_rest_helpers_nodeps.py) use the subprocess module to execute system commands including gcloud for authentication and curl for HTTPS requests. This is the intended mechanism for the skill to interact with the Google Cloud environment. The implementation includes safeguards such as using shutil.which to locate binaries and writing sensitive headers (containing OAuth2 tokens) to a temporary file with restricted permissions (0600) to ensure they do not appear in process arguments.
[EXTERNAL_DOWNLOADS]: The skill makes authenticated network requests to various official Google Cloud API endpoints, such as bigquery.googleapis.com, cloudresourcemanager.googleapis.com, and storageinsights.googleapis.com. These requests are necessary to retrieve the telemetry signals required for the security assessment. All domains are well-known, trusted service providers.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data including bucket names, resource tags, and dataset descriptions retrieved from the cloud environment. This data is interpolated into BigQuery SQL queries and assessment reports. To mitigate injection risks, the skill implements strict input validation in scripts/validation.py, using regular expressions for dataset and bucket names and blocking characters (like backticks) that could be used for SQL injection in project IDs.

gcs-security-assessment