cortex-agents-with-custom-tools-tutorial

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill guides the user through environment setup using the ACCOUNTADMIN role. While high-privilege, this is the standard and required practice for administrative tutorials that involve creating account-level objects like Cortex Search services and Agent configurations.
  • [SAFE]: All operations are performed within the Snowflake SQL and Snowpark (Python) environment. There are no external network calls, file system accesses outside the session, or remote script downloads detected.
  • [SAFE]: Security best practices for the Snowflake platform are reinforced, specifically the use of EXECUTE AS CALLER for stored procedures to ensure that the agent operates within the caller's specific RBAC permissions.
  • [SAFE]: No evidence of prompt injection, obfuscation, or hardcoded credentials was found in the tutorial instructions or the sample code provided.
  • [SAFE]: The skill demonstrates a legitimate vulnerability surface for indirect prompt injection inherent to RAG (Retrieval-Augmented Generation) architectures by processing documents from a search service. This is noted as a fundamental architectural property of the technology being taught rather than a malicious implementation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 10:41 PM