Pretty sure we can make LLM agents to transform declarative inputs to agentic action.