awesome-ai-harness README

The project's README from github.com/weiwei966/awesome-ai-harness. As one markdown file: README.md.

Awesome AI Harness Awesome

The model is the engine. The harness is the car.

A curated list of resources on harness engineering — the discipline of building the runtime scaffolding that turns a language model into a working agent.

中文版 →

Why this list

Most "awesome agent" lists catalogue frameworks. This one is about the layer underneath: the concrete engineering decisions that determine whether an agent works or flails.

A frontier model dropped into a bad harness behaves like a bad model. It loops, forgets, calls the wrong tool, burns its context on noise, and gives up. The same model in a good harness plans, recovers from errors, and finishes. The gap between those two outcomes is harness engineering, and almost none of it is in the model weights.

The craft is young, the knowledge is scattered across engineering blogs and leaked system prompts, and there is no textbook. This is an attempt at a map.

Scope note: this list covers the inference-time harness. Harnesses that double as RL training environments (environment hubs, trajectory generation, reward design) are a sibling discipline we deliberately leave out.

What is a harness?

Everything between the user's intent and the model's next token:

flowchart TB
    U([User intent]) --> H
    subgraph H[The Harness]
        direction TB
        SP[System prompt<br/>role, rules, environment]
        CTX[Context assembly<br/>budget · retrieval · compaction]
        TOOL[Tool layer<br/>definitions · results · errors]
        LOOP[Agent loop<br/>plan → act → observe → repeat]
        MEM[Memory &amp; state<br/>files · sessions · scratchpads]
        GATE[Permissions &amp; sandbox<br/>allowlists · HITL gates]
    end
    H <--> M{{Model}}
    H --> W([Effects on the world])

The model is a stateless function. Every bit of statefulness, safety, memory, and competence an "agent" appears to have was engineered into the loop around it.

Contents

Foundations

Start here if you are new to the field.

Terminology

The field's vocabulary is still settling; these pin it down.

Design patterns & surveys

Context Engineering

The context window is the agent's entire working memory, and it is a scarce, actively-managed resource. This is where most harness quality lives.

Tool Design & Agent-Computer Interfaces

Tools are the agent's API to reality. Their names, schemas, return values, and error messages are prompt engineering by another name — and the interface design is worth more than a model upgrade.

The Agent Loop

Memory, State & Session Resume

Multi-Agent & Orchestration

Sandboxing & Runtime Isolation

If your harness can run commands, it can run bad commands. The 2025-26 convergence: both major labs independently arrived at the same two-plane design — filesystem isolation plus a domain-allowlist egress proxy.

Safety, Permissions & Human-in-the-Loop

awesome-ai-harness (weiwei966/awesome-ai-harness)

Searching Database…

153 ·GitHub
Awesome-ai-harness
What is this?A curated reading list on 'harness engineering' — the runtime scaffolding (context management, tool design, agent loops, memory, sandboxing, evaluation) that turns a raw language model into a working agent, in English and Chinese.
Is this a tool to install, or just a reading list?It is a pure markdown awesome-list of links to papers, engineering blog posts, and talks, organized into topic sections — there is no code to run.
How is the list organized?It's split into sections like Foundations, Context Engineering, Tool Design, The Agent Loop, Memory/State/Session Resume, Multi-Agent & Orchestration, Sandboxing & Runtime Isolation, Safety & Permissions, Protocols & Interop, Skills, and Evaluation & Observability, with a Foundations section flagged for newcomers.
Does the list cover RL training environments too?No, it deliberately excludes that. It scopes itself to the inference-time harness and treats RL training environments (environment hubs, trajectory generation, reward design) as a separate discipline outside its coverage.
Who is this reading list for?Developers building coding agents, AI assistants, or agent runtimes who need to understand context budgeting, tool interface design, memory, sandboxing, and evaluation beyond just swapping in a bigger model.
What exactly is a 'harness'?Per the README, it's everything between the user's intent and the model's next token: the loop, tools, memory, safety checks, and context plumbing engineered around a stateless model to make it act like a competent agent.
Why is a 'harness' different from an 'agent framework'?The list cites Arize's distinction that a harness owns the agent loop itself, while a framework just lends you parts to build one — a harness is the runtime, not a library of components.
Is there a Chinese version of this list?Yes, the README links a 中文版 (Chinese edition) and also includes Chinese-translated resources, such as a full Chinese translation of the context-engineering survey.
What's the one resource to read first?The README points to Lilian Weng's 'Harness Engineering for Self-Improvement' as the most complete map of the discipline, saying if you read only one thing on the list, read that.
What are similar projects?awesome-harness-engineering, awesome-harness-engineering, Agent-Learning-Hub, hive, harness-engineering, awesome-chatgpt-zh, GLM-5.3-Flash-J-Space-Capability-Realization-Report, ai-system-design
Which topics is it in?sandboxing, harness-engineering, agent-harness, context-engineering, evaluation, memory, orchestration, prompt-engineering, rag, agentic-ai, anthropic, multi-agent
Creator
weiwei966
Size
0.1 MB
Stars
153
Updated
CommitAug 19
Age
2mo
Commits · 90 Days