THE SEVEN-LAYER ETCLOVG TAXONOMY

HOW THE LAYERS WORK TOGETHER The layers are not a stack of separate boxes that never touch. They constantly pass information to one another. Execution asks Tooling to run a tool, Context decides what the model sees about the result, Lifecycle watches how long the run has lasted, Observability records each step, Verification scores the outcome, and Governance checks permissions along the way. Think of an orchestra. Each section plays its own part, but the music only works when all sections stay in time with each other. Because the layers depend on each other, a weakness in one can look like a weakness in another. For example, an agent might give a wrong answer because the Context layer left out a key document, but the symptom shows up as a Verification failure when a test catches the wrong answer. The skill of a harness engineer is tracing a symptom back to the layer that caused it. A useful habit is to ask, for any problem, which layer is the root cause, which layers merely show the symptom, and which layer could have caught it earlier. You will practice this habit often, and the assistant prompts in this chapter are built around it. Chapter 8 will show the actual components that implement these layers in a typical production harness.