gcs-security-assessment
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes system binaries gcloud and curl via Python's subprocess module to fetch authentication tokens and perform REST API calls to Google Cloud services. This behavior is essential for the skill's primary function of assessing cloud security posture. Evidence: scripts/cloud_rest_helpers_nodeps.py uses subprocess.run to call gcloud auth application-default print-access-token and curl for HTTPS requests.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes data ingested from external sources, specifically BigQuery datasets containing GCS bucket and object metadata, which could potentially contain malicious instructions intended to influence the agent's behavior. Ingestion points: Metadata fetched via scripts/fetch_bucket_telemetry.py and scripts/fetch_object_telemetry.py from BigQuery tables. Boundary markers: The skill instructions provide strict phase-based workflows and require grounding all findings in verified telemetry signals. Capability inventory: The skill can execute shell commands (running its own telemetry scripts) and interact with Google Cloud APIs. Sanitization: Input validation is implemented in scripts/validation.py using regular expressions to sanitize project IDs, dataset names, and bucket names before they are used in API calls or SQL queries.
Audit Metadata