Agent Harness 的剖析
TLDR: Agent = Model + Harness. Harness engineering is how we build systems around models to turn them into work engines. The model contains the intelligence and the harness makes that intelligence useful. We define what a harness is and derive the core components today's and tomorrow's agents need.
TLDR: Agent = Model + Harness。Harness engineering 是我们围绕 models 构建系统以将它们转变为工作引擎的方式。model 包含智能，而 harness 使这种智能变得有用。我们定义了 harness 是什么，并推导出当今和未来的 agents 所需的核心组件。
Can Someone Please Define a "Harness"?
有人能定义一下“Harness”吗？
Agent = Model + Harness
Agent = Model + Harness
If you're not the model, you're the harness.
如果你不是 model，你就是 harness。
A harness is every piece of code, configuration, and execution logic that isn't the model itself. A raw model is not an agent. But it becomes one when a harness gives it things like state, tool execution, feedback loops, and enforceable constraints.
一个 harness 是除了模型本身之外的所有代码、配置和执行逻辑。一个原始模型不是代理。但当一个 harness 赋予它状态、工具执行、反馈循环和可强制执行的约束等功能时，它就变成了一个。
Concretely, a harness includes things like:
具体来说，harness 包括诸如：
- System Prompts
- System Prompts
- Tools, Skills, MCPs + and their descriptions
- Tools、Skills、MCPs + 及其描述
- Bundled Infrastructure (filesystem, sandbox, browser)
- Bundled Infrastructure (filesystem, sandbox, browser)
- Orchestration Logic (subagent spawning, handoffs, model routing)
- Orchestration Logic（subagent spawning、handoffs、model routing）
- Hooks/Middleware for deterministic execution (compaction, continuation, lint checks)
- Hooks/Middleware 用于确定性执行（compaction、continuation、lint checks）
There are many messy ways to split the boundaries of an agent system between the model and the harness. But in my opinion, this is the cleanest definition because it forces us to think about designing systems around model intelligence.
划分 agent 系统中模型与 harness 边界有许多混乱的方式。但在我看来，这是最清晰的定义，因为它迫使我们思考围绕模型智能设计系统。
The rest of this post walks through core harness components and derives why each piece exists working backwards from the core primitive of a model.
本文其余部分将逐步介绍核心 harness 组件，并从模型的核心原语反推推导出每个部分的存在原因。
