revfactory/harness
A meta-skill that designs domain-specific agent teams, defines specialized agents, and generates the skills they use.
HTML · ★ 7,900 · Fork 1,071 · 在 GitHub 查看
在手机上浏览源码、订阅每日 AI 趋势日报与项目架构深度解读:获取 CopoHub 客户端(iOS & HarmonyOS)→
GitHub Trending 上榜记录
revfactory/harness 已在 GitHub Trending 每日榜单上榜 9 天,最高排名第 5,首次上榜 2026-05-28,最近一次上榜 2026-06-25。
| 日期 | 排名 | Star | 当日增长 |
|---|---|---|---|
| 2026-06-25 | #6 | 7,900 | +277 |
| 2026-06-24 | #6 | 7,697 | +274 |
| 2026-06-23 | #10 | 7,419 | +123 |
| 2026-06-02 | #10 | 5,393 | +524 |
| 2026-06-01 | #8 | 5,100 | +527 |
| 2026-05-31 | #5 | 4,562 | +318 |
| 2026-05-30 | #5 | 4,222 | +313 |
| 2026-05-29 | #10 | 4,060 | +65 |
| 2026-05-28 | #10 | 3,846 | +190 |
同期上榜的项目
- google-labs-code/design.md – A format specification for describing a visual identity to coding agents. DESIGN.md gives agents a persistent, structured understanding of a design system.
- calesthio/OpenMontage – World's first open-source, agentic video production system. 12 production pipelines, 100+ tools, 700+ agent skill and production-knowledge files. Turn your AI coding assistant into a full video production studio.
- xbtlin/ai-berkshire – AI 时代的伯克希尔:基于 Claude Code / Codex 的价值投资研究框架。巴菲特·芒格·段永平·李录四大师方法论 + 多Agent并行研究。| AI-era Berkshire: a value investing research framework built for Claude Code / Codex. 4 masters' methodologies + multi-agent adversarial analysis.
- mauriceboe/TREK – A self-hosted travel/trip planner with real-time collaboration, interactive maps, PWA support, SSO, budgets, packing lists, and more.
- apple/container – A tool for creating and running Linux containers using lightweight virtual machines on a Mac. It is written in Swift, and optimized for Apple silicon.
- JCodesMore/ai-website-cloner-template – Clone any website with one command using AI coding agents
Harness 项目深度剖析与商业评估报告
Harness 项目深度剖析与商业评估报告
1. 核心定位与商业落地价值
1.1 根本痛点解决 Harness 解决的是 LLM Agent 系统架构设计的非确定性和高成本问题。当前 Claude Code 等平台虽然提供了 Agent 能力,但团队架构设计(选择何种协作模式、定义 Agent 角色、编写技能文件)完全依赖工程师的直觉和经验,导致: 每次项目启动都需要重复设计,缺乏可复用的架构模板 团队协作模式不一致,Pipeline、Fan-out/Fan-in 等模式混用导致混乱 技能文件编写成本高,每个 Agent 需要手工编写 SKILL.md 和 reference 文档
Harness 将这一过程从 手工操作提升到元编程层面,通过输入领域描述即可自动生成完整的 Agent 团队定义和技能文件。
1.2 企业级商业落地场景
| 场景 | 价值量化 | 适用行业 | |------|---------|---------| | 代码审查流水线 | 并行架构/安全/性能审查,减少 60% 人工审查时间 | 金融、医疗、合规 | | 文档生成工厂 | API 文档、用户手册自动生成,降低 80% 文档维护成本 | SaaS、云服务 | | 营销内容工厂 | 多语言、多渠道内容生成,提升 3x 内容产出速度 | 广告、电商 | | 数据管道设计 | 自动生成 ETL 逻辑和 schema 设计,减少 50% 数据工程师工作量 | 数据中台、BI |
1.3 竞品差异化与护城河
vs Archon(coleam00/Archon):Archon 解决运行时配置的确定性,Harness 解决团队架构设计。两者是 互补而非竞争 关系。Archon 关注“如何执行”,Harness 关注“谁在执行、如何协作”。 vs LangGraph:LangGraph 是状态图编排,适合长时间运行、状态可恢复的复杂流程;Harness 是 Claude Code 原生、轻量级团队设计,适合快速迭代的 Agent 团队。 vs wshobson/agents:后者是 Agent/Skill 的“零件目录”,Harness 是“工厂”。Harness 可以吸收 wshobson 的零件作为子组件,形成 工厂+供应链 的生态协同。
在 README 中展示上榜徽章
[](https://github.fq6825.top/repo/revfactory/harness)