databricks-zerobus-ingest
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the
databricks-sdk,databricks-zerobus-ingest-sdk, andgrpcio-toolslibraries via pip on the Databricks execution environment. These are legitimate packages corresponding to the vendor's services and well-known communication utilities. - [COMMAND_EXECUTION]: The workflow requires executing Python code on Databricks clusters using the environment's provided MCP tools (
run_python_file_on_databricksandexecute_databricks_command). This is the intended behavior for deploying and testing ingestion clients within the lakehouse environment. - [CREDENTIALS_UNSAFE]: While the skill manages Service Principal credentials, it explicitly instructs against hardcoding them, instead recommending environment variables and utilizing descriptive placeholders in documentation.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it generates code to ingest external records and processes Unity Catalog table metadata. However, it incorporates specific workflow steps for validation and utilizes structured Protobuf schemas which mitigate common data-driven injection risks (Severity: LOW).
Audit Metadata