The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses the Databricks CLI for authentication, resource configuration, and deployment tasks (e.g., databricks bundle deploy, databricks bundle run). These are standard operations for managing the Databricks platform.
[EXTERNAL_DOWNLOADS]: Fetches MLflow model artifacts from the Databricks Model Registry using the mlflow.artifacts.download_artifacts API. This is a trusted internal source within the user's workspace context.
[REMOTE_CODE_EXECUTION]: The skill facilitates the download of Python scripts from a remote Model Serving endpoint to a local directory (./original_mlflow_model/code/) and subsequently executes them in a new app environment using uv run start-app and databricks bundle run.
[PROMPT_INJECTION]: The skill identifies an indirect injection surface by processing and executing code artifacts retrieved from a model registry.
Ingestion points: Python files are downloaded from the remote Model Registry into original_mlflow_model/code/ for migration.
Boundary markers: No specific delimiters or safety instructions are used to isolate the downloaded code from the migration agent's context.
Capability inventory: The skill is capable of executing the migrated code locally and in a production app environment via uv and databricks bundle commands.
Sanitization: The migration instructions do not include steps for sanitizing or auditing the downloaded code before it is executed.

migrate-from-model-serving