The Agent Skills Directory

[SAFE]: The skill demonstrates secure handling of authentication credentials and tokens. It correctly instructs users to utilize the Databricks SDK configuration and environment variables for OAuth flows, preventing the exposure of hardcoded secrets.
[COMMAND_EXECUTION]: The documentation includes standard Databricks CLI commands for application management, such as creating apps, importing workspace files, and managing deployments. These are routine administrative actions for the target platform.
[EXTERNAL_DOWNLOADS]: The skill references standard Python packages (e.g., psycopg2-binary, asyncpg) for database integration. These are well-known, trusted libraries used for PostgreSQL connectivity.
[PROMPT_INJECTION]: The provided example code includes patterns for processing external data and user input through Large Language Models, which creates a surface for potential indirect prompt injection.
Ingestion points: User-provided chat inputs in examples/fm-minimal-chat.py and model-generated responses in examples/fm-structured-outputs.py.
Boundary markers: The skill employs structured system prompts to define JSON schemas and output constraints, though explicit delimiters for user input text are not featured in the examples.
Capability inventory: The code can interact with SQL warehouses and foundation model endpoints using the Databricks Python SDK.
Sanitization: Robust JSON parsing logic is implemented in examples/fm-structured-outputs.py to handle malformed or potentially adversarial content in model responses.

databricks-app-python