databricks-streaming-guardian
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill incorporates a
PreToolUsehook (hooks/streaming-guard-hook.py) and helper scripts (scripts/pre-optimize-check.sh,scripts/recover-streaming-source.py) that execute shell commands. These components are designed to interact with thedatabricksCLI andjqto safely query workspace state and table properties. - [REMOTE_CODE_EXECUTION]: The hook and associated scripts make network requests to the official Databricks Statement Execution API. These requests are used to retrieve metadata about active streaming queries and table versions, operating within the user's existing authenticated session context.
- [PROMPT_INJECTION]: The
eval-spec.yamlfile includes text patterns associated with prompt injection in its test cases. These are used exclusively for automated testing to ensure the agent resists such attacks and do not represent a vulnerability in the skill itself.
Audit Metadata