The Agent Skills Directory

[COMMAND_EXECUTION]: The Python scripts execute_m.py and preview_partition.py use subprocess.run to call the Azure CLI (az) for authentication tokens and the fab CLI for retrieving model definitions. This behavior is standard for tools designed to automate Fabric operations.
[EXTERNAL_DOWNLOADS]: Documentation and scripts reference the use of the pyarrow library, a well-known and trusted package for processing Arrow-formatted data returned by the Fabric API.
[REMOTE_CODE_EXECUTION]: The skill facilitates the execution of Power Query M code via the official Microsoft Fabric executeQuery endpoint. This is the core functionality intended for testing and validating data transformations.
[DYNAMIC_EXECUTION]: The preview_partition.py script employs __import__ and sys.path modifications to load its companion module execute_m.py at runtime. While this is a form of dynamic loading, it is used here to reference local project files rather than untrusted external code.
[PROMPT_INJECTION]: The skill ingests semantic model definitions (TMDL) from external Fabric workspaces. This represents an indirect prompt injection surface where a compromised model definition could potentially contain malicious M code or instructions.
Ingestion points: preview_partition.py retrieves TMDL content via the fab get command.
Boundary markers: Absent; the script parses the raw TMDL payload directly.
Capability inventory: Includes network requests to Fabric APIs (execute_m.py) and local command execution via subprocess.
Sanitization: No sanitization or validation is performed on the M expressions extracted from the TMDL files before execution.

power-query