Agent Systems with Harness Engineering: A Systematic Survey

Abstract

Constructing agents that operate effectively in open environments has remained a central goal of artificial intelligence research. Large language models have recently shifted the design space for such agents, yet even strong models, when placed in an inadequately supported runtime, fail systematically on extended tasks: they lose continuity of plans, call tools with poorly formed arguments, and continue acting without consulting the most recent state of the world. Such failures point less to a deficit in model training than to an architectural mismatch: the model exposes a one-shot textual interface, whereas realistic tasks demand persistent state and iterative refinement. In this survey, we formalize the supporting infrastructure as the harness and introduce harness engineering as the joint optimization of both the harness and the model it supports. The relationship between the two is inherently synergistic: a well-designed harness unlocks latent model potential through structured memory, error recovery, and adaptive context management, while a capable model enables the harness to implement sophisticated workflows that would be infeasible with a weaker reasoner. We present a structured taxonomy of harness components, spanning agent workflows, memory systems, skill libraries, and multi-agent orchestration, and analyze how each contributes to system-level performance, distilling design principles and architectural trade-offs from current practice. We further review optimization strategies from both the scaffold side and the model side, and summarize evaluation benchmarks across software engineering, deep research, tool use, computer use, and scientific discovery. By shifting focus from what agents can accomplish to how the supporting infrastructure enables them to do so, this survey aims to serve as a practical reference for building reliable, scalable, and controllable agent systems through principled harness engineering. To support ongoing research, we also maintain a curated GitHub repository of essential harness engineering resources and supplementary materials, accessible at the website https://github.com/RUCAIBox/awesome-agent-harness.

Keywords

Harness engineering / Agent systems / Large language models

Rights & permissions

Higher Education Press 2026

Just Accepted

This article has successfully passed peer review and final editorial review, and will soon enter typesetting, proofreading and other publishing processes. The currently displayed version is the accepted final manuscript. The officially published version will be updated with format, DOI and citation information upon launch. We recommend that you pay attention to subsequent journal notifications and preferentially cite the officially published version. Thank you for your support and cooperation.