A Field Guide to Agent Harness Engineering - The Architecture Behind Reliable AI Agents

Abstract

Why I am writing this: For most of the last three years, the implicit thesis of agentic AI has been simple: a better model produces a better agent. Buy the frontier model, write a good prompt, wire up some tools, and reliability follows. Between March and June 2026, a cluster of papers landed that, read together, argue something close to the opposite. The binding constraint on real-world agent reliability is increasingly not the model. It is the harness — the runtime system that wraps the model and governs how it observes, acts, remembers, recovers, and is verified. This is not a vocabulary game. It changes how you staff a team, where you spend compute budget, what you put in a benchmark, and what you can honestly promise a board. Below I work through seven papers, group them into a coherent picture, and end with what I think it means in practice. References are listed in full at the end so the claims here are traceable to their sources rather than to me. A note on framing for my PhD students: I have deliberately kept the formal objects (the labeled-transition-system semantics, the component algebra) close to how the authors state them, because the precision is the point. For my consulting readers, each section closes with a 'so what' that you can act on without the math.