Code-oriented LLMs and tool-use models for agent execution workflows. Used to evaluate planning, refactoring, and structured action outputs.