The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The tune.py script dynamically loads and executes Python code from a local file named udf_setup.py if it exists in the project directory. This allows for arbitrary code execution within the agent's environment.
[EXTERNAL_DOWNLOADS]: The skill suggests installing the uv package manager using a remote script from astral.sh piped directly into the shell. While originating from a well-known service, this installation pattern is a common vector for remote code execution.
[COMMAND_EXECUTION]: Multiple scripts including discover.py, env_setup.py, and init_run.py use the subprocess module to execute system commands such as the Databricks CLI, Git, and Python.
[PROMPT_INJECTION]: The skill processes SQL queries from project files and executes them on Databricks. This presents an indirect prompt injection surface where malicious instructions could be embedded in data files.
Ingestion points: SQL query files or inline strings provided via the --query or @file arguments.
Boundary markers: Absent; the skill treats processed SQL as executable code.
Capability inventory: Shell command execution via subprocess.run and Spark SQL operations through the Databricks connection.
Sanitization: No explicit sanitization or validation of the SQL content is performed prior to its execution by the Spark engine.

databricks-sql-autotuner