The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses spark.sql() within several files (SKILL.md, scripts/register_prompts.py, references/loading-patterns.md) to manage prompt storage. This is the standard method for interacting with Unity Catalog tables in a Databricks environment and is used for its intended purpose of configuration management.
[PROMPT_INJECTION]: The skill architecture involves loading prompts from external storage at runtime, which is a common pattern for agent configuration.
Ingestion points: Prompts are retrieved from the agent_config table in Unity Catalog.
Boundary markers: Not explicitly enforced in the logic, as the skill assumes governed access to the configuration table.
Capability inventory: The skill utilizes spark.sql for data retrieval and mlflow for artifact versioning.
Sanitization: The implementation includes explicit logic to escape single quotes (.replace("'", "''")) to prevent SQL injection during prompt registration.
[EXTERNAL_DOWNLOADS]: The skill references documentation and GitHub repositories owned by databricks-solutions and mlflow. These are trusted organizations, and the references are documented neutrally as legitimate technical dependencies.

prompt-registry-patterns