gcp-audit-logs
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses gcloud, bq, and gsutil CLI tools to modify organization and project-level IAM policies, create log sinks, and manage storage resources. These are standard operations for GCP infrastructure management.
- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection. 1. Ingestion points: GCP log entries are ingested into the context via BigQuery SQL queries and gcloud logging read commands defined in SKILL.md. 2. Boundary markers: The provided analysis templates do not include markers to distinguish between data and instructions. 3. Capability inventory: The skill utilizes gcloud, bq, gsutil, and terraform for resource management and log retrieval. 4. Sanitization: No sanitization or filtering logic is present in the provided query templates.
Audit Metadata