Harness, Scaffold, and the AI Agent Terms Worth Getting Right

The vocabulary around AI agents is evolving rapidly, leading to confusion over terms like 'harness' and 'scaffold.' This glossary defines a model as the core LLM, while scaffolding is the behavior-defining layer containing system prompts and tool descriptions. The harness is the execution layer that calls the model, handles tool use, and manages the operational loop. An agent is the complete system, combining the model with the scaffolding and harness to enable autonomous action. Other related concepts such as context engineering, policy, and tool use are also explained to provide a practical mental model for practitioners.