Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add nilecui/harness-zh-release --skill harness-zhgit clone --depth 1 https://github.com/nilecui/harness-zh-releaseWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nilecui/harness-zh-release/harness-zh)<a href="https://agentmods.dev/skills/nilecui/harness-zh-release/harness-zh"><img src="https://agentmods.dev/badge/skills/nilecui/harness-zh-release/harness-zh/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nilecui/harness-zh-release/harness-zh"><img src="https://agentmods.dev/badge/skills/nilecui/harness-zh-release/harness-zh.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00203 | $0.05966 |
| Opus 5 | $0.00102 | $0.02983 |
| Sonnet 5 | $0.00041 | $0.01193 |
| Haiku 4.5 | $0.00020 | $0.00597 |
Grade A, and why
harness-zh scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 357 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Harness — Agent Team & Skill Architect
为特定领域/项目构建 Harness，定义各个 Agent 的角色，并生成 Agent 所使用的 Skill 的元 Skill（meta skill）。
核心原则：
- 生成 Agent 定义（
.claude/agents/）与 Skill（.claude/skills/）。 - 将 Agent 团队（Agent Team）作为默认执行模式。
- 在 CLAUDE.md 中注册 Harness 指针（pointer）。 —— 仅记录最少量的指针（触发规则 + 变更历史），以便在新会话中自动触发编排器（orchestrator）Skill。
- Harness 不是固定物，而是不断进化的系统。 —— 每次执行后都要吸收反馈，持续更新 Agent、Skill 与 CLAUDE.md。
工作流（Workflow）
Phase 0: 现状审计（Audit）
当 Harness Skill 被触发时，首先确认既有 Harness 的现状。
-
读取
项目/.claude/agents/、项目/.claude/skills/、项目/CLAUDE.md -
根据现状分支执行模式：
- 全新构建：Agent/Skill 目录不存在或为空 → 从 Phase 1 开始完整执行
- 既有扩展：已有 Harness，并需要追加新 Agent/Skill → 按下方 Phase 选择矩阵仅执行必要 Phase
- 运维/维护：对既有 Harness 的审计·修改·同步请求 → 跳转至 Phase 7-5 运维/维护工作流
既有扩展时的 Phase 选择矩阵：
变更类型 Phase 1 Phase 2 Phase 3 Phase 4 Phase 5 Phase 6 新增 Agent 跳过（复用 Phase 0 结果） 仅决定编排位置 必需 需要专属 Skill 时 修改编排器 必需 新增/修改 Skill 跳过 跳过 跳过 必需 连接关系变化时 必需 架构变更 跳过 必需 仅受影响 Agent 仅受影响 Skill 必需 必需 -
将既有 Agent/Skill 列表与 CLAUDE.md 记录进行对照，检测不一致（drift）
-
将审计结果汇总汇报给用户，并确认执行计划
Phase 1: 领域分析（Domain Analysis）
- 从用户请求中把握领域/项目
- 识别核心作业类型（生成、校验、编辑、分析等）
- 基于 Phase 0 的审计结果，分析与既有 Agent/Skill 的冲突/重复
- 探索项目代码库 —— 把握技术栈、数据模型、主要模块
- 检测用户熟练度 —— 通过对话中的上下文线索判断其技术水平，据此调节后续沟通语气。对编程经验较少的用户，避免不加解释地使用专业术语。
Phase 2: 团队架构设计（Team Architecture Design）
2-1. 选择执行模式
Agent 团队是最高优先级的默认值。 当有 2 个以上 Agent 协作时，必须首先考察是否采用 Agent 团队。团队成员之间通过直接通信（SendMessage）与共享任务列表（TaskCreate）自我协调，通过发现共享、冲突讨论、遗漏补全来提升结果质量。
| 模式 | 何时使用 | 特性 |
|---|---|---|
| Agent 团队（默认） | 2 人以上协作、需要实时协调·反馈交换、相互引用中间产物 | 通过 TeamCreate + SendMessage + TaskCreate 自我协调 |
| 子 Agent（Sub Agent）（备选） | 单 Agent 作业、只需向主体返回结果即可、团队通信开销过大时 | 直接调用 Agent 工具，使用 run_in_background 并行 |
| 混合（Hybrid） | 各 Phase 特性不同时 —— 如：并行收集（子 Agent）→ 基于共识的整合（团队） | 按 Phase 混合团队/子 Agent |
决策顺序：
- 首先考察是否可按 Agent 团队设计 —— 2 人以上即为默认
- 仅当结构上不需要团队通信（只需结果传递）、且团队开销大于收益时，才选择子 Agent
- 各 Phase 特性差异明显时考虑混合 —— 将各 Phase 的执行模式在编排器中明示
What ships with it
6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 357 lines · 203 tokens per session scan A 0508031ee972
harness-zh is a skill published in the GitHub repository nilecui/harness-zh-release (4 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 203 tokens to every session and 5,966 once invoked, about $0.0010 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
evolve
A harness-improvement skill for reviewing how an agent setup performed and feeding lessons back into its agents, skills, and orchestration.
docx
Instructions for safely reading, creating, and editing Microsoft Word documents and templates, including tables, images, comments, and tracked changes.
pptx
Instructions for reading, creating, and editing PowerPoint presentation files and templates, including .pptx, .potx, and older .ppt files. They cover text, tables, images, charts, layouts, and speaker notes.
A guide for working with PDF files, including reading, extracting content, creating, editing, and checking documents.
great_cto
Use when the CTO describes a feature, task, or project goal. Orchestrates the full SDLC pipeline automatically based on project type.
anti-patterns
Catalogue of known SDLC anti-patterns that greatcto agents must actively reject when reviewing architecture, plans, code, or post-mortems. Used by architect (pre-impl), pm (planning), senior-dev (impl), l3-support (post-incident).