spark-consumption-cli

Pass

Audited by Gen Agent Trust Hub on Mar 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • Command Execution: The skill utilizes the Azure CLI (az rest) to interact with the Microsoft Fabric API. This is the standard mechanism for CLI-based management of Fabric workspaces and items, allowing the agent to perform administrative and discovery tasks.\n- Remote Code Execution: It facilitates the execution of Spark and PySpark code via Livy sessions. This is a fundamental feature of the skill, enabling advanced data exploration and analytics on remote compute clusters.\n- Local File Interaction: Temporary JSON files are created in the /tmp directory to prepare request bodies for API communication. This is a common and practical method for handling structured data in shell-based workflows.\n- Indirect Prompt Injection Surface: Because the skill processes data from external lakehouses, it possesses an ingestion point for potentially untrusted content.\n
  • Ingestion points: Reads data from lakehouse tables and unstructured files (SKILL.md).\n
  • Boundary markers: Spark code is encapsulated within structured JSON objects sent to the Livy REST API.\n
  • Capability inventory: Execution of PySpark/Spark SQL code, OneLake data access, and session lifecycle management (SKILL.md).\n
  • Sanitization: Relies on platform-level security (RBAC) and the internal safety filters of the underlying language model to ensure generated code is safe for execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 19, 2026, 12:32 PM