databricks-data-discovery

Pass

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the databricks experimental genie ask command to process data queries. This is the intended primary function of the skill and uses the official vendor-provided CLI tool.
  • [PROMPT_INJECTION]: The skill includes an attack surface for indirect prompt injection as it processes user-provided natural language and passes it to the Databricks Genie service.
  • Ingestion points: Natural language questions provided by the user are passed directly into shell commands as arguments in SKILL.md.
  • Boundary markers: Example commands utilize double quotes to encapsulate the user's query string, providing a basic delimiter.
  • Capability inventory: The skill possesses the ability to execute shell commands via the databricks CLI, which can interact with the organization's Unity Catalog metadata and data metrics.
  • Sanitization: The instructions do not specify explicit sanitization, relying on the agent's command execution safety and the CLI's internal handling of the query string.
  • [SAFE]: The skill is authored by 'databricks' and exclusively uses the official Databricks CLI and its associated services (Genie/Unity Catalog). There are no signs of malicious obfuscation, persistence, or data exfiltration to unauthorized destinations.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 25, 2026, 07:07 PM
Security Audit — agent-trust-hub — databricks-data-discovery