bigquery-ai-ml

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill provides detailed reference material for BigQuery's AI and ML functions. A thorough review found no evidence of malicious instructions, obfuscation, or unauthorized data access. The examples use public datasets and standard BigQuery syntax.
  • [INDIRECT_PROMPT_INJECTION]: The skill documents functions that analyze unstructured data from database tables (e.g., user reviews), which is an inherent surface for indirect prompt injection in LLM-based analytics.
  • Ingestion points: The input argument in AI.AGG (references/ai_agg.md) and prompt in AI.GENERATE (references/ai_generate.md) accept text data from arbitrary table columns.
  • Boundary markers: The provided examples do not explicitly demonstrate the use of delimiters or specific instructions to isolate data from commands.
  • Capability inventory: The functions enable LLM-driven aggregation, classification, and text generation via BigQuery and Vertex AI.
  • Sanitization: No input validation or sanitization methods are described in the documentation for these functions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 05:41 AM
Security Audit — agent-trust-hub — bigquery-ai-ml