The Agent Skills Directory

[SAFE]: The skill provides legitimate instructions for interacting with Databricks workspace environments using an MCP tool. The behavior described is consistent with the skill's stated purpose and author context.- [DATA_EXFILTRATION]: The skill references the sensitive configuration file path ~/.databrickscfg. It instructs the agent to use a tool to manage this configuration instead of performing direct file manipulation.- [PROMPT_INJECTION]: The skill ingests untrusted data from external Databricks workspace environments, which constitutes an indirect prompt injection surface.
Ingestion points: Workspace profiles and host metadata returned by the manage_workspace tool.
Boundary markers: The skill does not define specific delimiters for tool output.
Capability inventory: The manage_workspace tool enables listing, switching, and authenticating to remote workspaces.
Sanitization: The instructions do not specify sanitization or validation of the host or profile data before presentation.

databricks-config