Agent skill

Vibe Coding Harness

by Junliu1066 in Junliu1066/vibe-coding-kit

质检 Agent:检查 vibe-coding-kit 其他 Skill 的产出物是否符合治理规范. An agent skill from Junliu1066/vibe-coding-kit.

MITAuto-check passedProduct & Project Management

Install Vibe Coding Harness

skills CLI
$ npx skills add Junliu1066/vibe-coding-kit --skill vibe-coding-harness -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Junliu1066/vibe-coding-kit vibe-coding-harness --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Junliu1066/vibe-coding-kit.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/vibe-coding-harness .claude/skills/vibe-coding-harness && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
vibe-coding-harness
GitHub stars
109
Token cost
~1.3k tokens
SKILL.md length
277 words
Files
1
Skills in repo
6
Repo updated
First seen
Licence
MIT

At a glance

质检 Agent:检查 vibe-coding-kit 其他 Skill 的产出物是否符合治理规范. An agent skill from Junliu1066/vibe-coding-kit.

  • Works in 4 steps: :确定检查范围 → :加载规则 → :逐项检查 → …
  • Tasks that involve PRD writing
  • SKILL.md covers 检查流程, 修复与重试, 使用方式 and 速查:各 Skill 的检查要点, plus 1 more section
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Vibe Coding Harness is an agent skill from Junliu1066/vibe-coding-kit. 质检 Agent:检查 vibe-coding-kit 其他 Skill 的产出物是否符合治理规范。 当用户说"检查一下产出"、"帮我看看合不合规"、"质检"、"验证 PRD"、"检查项目说明书", 或在任何 Skill 完成产出后,使用此 Skill 做合规检查。 它读取项目根目录的 harness.json,逐条校验产出文件,报告违规项和修复建议。 这是 vibe-coding-kit 套件的治理层(harness),确保所有产出物格式统一、质量达标。 它不止检查格式,还会做跨 Skill 一致性校验——比如 PRD 里的项目名和项目说明书里的项目名是不是一致。

Its SKILL.md is about 1.3k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Product & Project Management, covering PRD writing. The repository describes itself as: 不教你写代码,教你把"用 AI 写代码"彻底想明白 — 给非技术产品人的 vibe coding 规划工具包. The licence is MIT.

When your agent uses it

  • Tasks that involve PRD writing

Example prompts

  • “检查一下产出”
  • “帮我看看合不合规”
  • “验证 PRD”
  • “/vibe-coding-harness”

Workflow steps

4 steps, taken from the step headings in SKILL.md.

  1. :确定检查范围
  2. :加载规则
  3. :逐项检查
  4. :输出检查报告

What it can do on your machine

Read from SKILL.md and the folder at commit cb387af. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Vibe Coding Harness loads about 1.3k tokens when it runs. Until then it costs about 77 tokens; SKILL.md has 277 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~77
When it runs · the whole SKILL.md, loaded when a task matches
~1.3k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Junliu1066/vibe-coding-kit at commit cb387af, republished under its MIT licence (© Junliu1066). 277 words, ~1,290 tokens.

Download SKILL.mdSave it as .claude/skills/vibe-coding-harness/SKILL.md (or your agent's skills folder).
name
vibe-coding-harness
description
质检 Agent:检查 vibe-coding-kit 其他 Skill 的产出物是否符合治理规范。 当用户说"检查一下产出"、"帮我看看合不合规"、"质检"、"验证 PRD"、"检查项目说明书", 或在任何 Skill 完成产出后,使用此 Skill 做合规检查。 它读取项目根目录的 harness.json,逐条校验产出文件,报告违规项和修复建议。 这是 vibe-coding-kit 套件的治理层(harness),确保所有产出物格式统一、质量达标。 它不止检查格式,还会做跨 Skill 一致性校验——比如 PRD 里的项目名和项目说明书里的项目名是不是一致。

质检 Agent:产出物合规检查

这不是一个独立使用的 Skill,而是其他 Skill 的"守门员"——在任何 Skill 声称"完成"后,用它来验证产出物是否真正符合规范。

治理规则来自项目根目录的 harness.json(唯一真实来源)和 CLAUDE.md(人类可读宪法)。 本 Skill 把这些规则变成逐条可执行的检查项。


检查流程

Step 1:确定检查范围

先确认用户想检查什么:

  • 如果用户指定了 Skill(如"检查 PRD"),只检查那个 Skill 的产出
  • 如果用户说"全量检查"或没说具体,检查 docs/ 下所有已有产出物
  • 如果 docs/ 不存在或为空,直接报告"没有可检查的产出物"
Step 2:加载规则

读取项目根目录的 harness.json,提取:

  • 目标 Skill 的 produces(该产出什么文件)
  • 目标 Skill 的 required_sections(必须包含的章节)
  • 目标 Skill 的 field_requirements(字段级校验规则)
  • 目标 Skill 的 validation(内容级校验规则)
  • cross_skill_rules(跨 Skill 一致性规则,仅在多 Skill 产出共存时检查)
  • quality_gates.before_delivery(交付前必须通过的门)
  • workflow(流程定义:阶段、步骤、entry_gate/exit_gate、账本规则)—— 用于新增的流程一致性检查
  • progress_cards(项目进度卡规则:低上下文读取、总表/模块卡/任务卡同步)

并读取 docs/进度账本.md(流程状态来源)。若不存在,流程检查记为"未启用账本"。 如果项目存在 .dsu/progress/index.md,读取该文件并按需抽查相关模块卡和任务卡;如果不存在,项目进度卡检查记为"未启用"。

Step 3:逐项检查

按以下顺序逐项检查,每项给出通过/不通过/不适用:

3.0 流程一致性检查(最先做 · 这一版新增)

对照 harness.json.workflow 和 docs/进度账本.md,核对过程有没有跑偏(不只是产出物对不对):

  • 阶段按序:有没有跳过整个阶段而未推进(如账本显示在 S2,但 S1 出口门 ✗)。
  • 假完成:账本标「已完成」的阶段,其 exit_gate 是否真的过了(防止"账本说完成、产出物却没达标")。
  • 跳步留痕:被跳过的步骤是否都是 required:false、且在账本"跳步留痕"写了理由;required:true 的步骤被跳 = 🔴 违规。
  • 账本与产出物吻合:账本当前阶段/步骤,与 docs/ 下实际存在的产出物对得上。
  • 红线:若项目涉及"动钱 / 动别人隐私",survival 红线提示是否出现过。

输出:🔴 流程违规(如"S1 未过却已在 S2")/🟡 跳步未留痕、账本与进度说明书不一致/✅ 流程一致。

3.0.5 项目进度卡检查(开发期上下文治理)

如果项目启用了 .dsu/progress/,对照 harness.json.progress_cards 检查:

  • 总表是否存在:.dsu/progress/index.md 是否存在且非空。
  • 总表是否只放索引级信息:总表应只回答当前阶段、当前主线、完成 DU、进行中 DU、下一步、最大风险,不应堆模块实现细节。
  • 读取规则是否清楚:总表必须说明默认只读 index,按需读取模块卡,复盘才读任务卡。
  • 模块卡是否可下钻:DU 总进度表中的细节卡路径应指向 .dsu/progress/modules/duXX-xxx.md。
  • 任务 closeout 是否同步:最近任务更新后,应能在任务卡、模块卡、总表三处看到一致状态。
  • 状态枚举是否合规:只能使用 未开始 / 脚手架 / 进行中 / 已完成 / 阻塞 / 回滚 / 待重构。
  • 最小上下文是否可用:总表中的「下一轮 AI 最小上下文」应控制在 150 到 300 字,且不写历史流水账。

输出:🔴 缺少总表或状态冲突/🟡 总表过细、模块卡缺失、任务 closeout 未同步/✅ 项目进度卡可用。

3.1 文件存在检查

对照 produces,确认每个文件是否已创建、是否非空。

3.2 必须章节检查

对照 required_sections,在文件中搜索这些章节标题是否存在。

3.3 字段校验

对照 field_requirements:

  • 字段是否存在且非空
  • 是否匹配指定的正则(如版本号格式 v\d+\.\d+)
  • 是否满足最小长度/最小条目数
3.4 内容校验

对照 validation:

  • 字数是否在范围内
  • 是否包含必须出现的关键词
  • 是否为 checklist 格式(如果要求的话)
3.5 全局禁止项检查

对照 constitution.forbidden 逐条扫描:

  • 搜索 { } 配对、TODO、TBD 等占位符模式
  • 检查是否有大片英文正文(排除代码块)
  • 检查是否有疑似密钥/密码/token 的模式(长随机字符串、sk- 前缀等)
3.6 跨 Skill 一致性检查(仅当多 Skill 产出共存时)
  • CSR-001:对比 prd.md 的「产品结论」与 项目说明书.md 的「需求基准描述」
  • CSR-002:对比 PRD 的「形态」与项目说明书的「技术栈」
  • CSR-003:对比所有产出物中的项目名称
  • CSR-004:检查验收标准是否具体可验证
3.7 质量门检查

逐条对照 quality_gates.before_delivery。

Step 4:输出检查报告

按严重程度分组输出:

## 质检报告:{Skill 名称 / 全量检查}

### 🔴 错误(必须修复,否则产出不合格)
| # | 文件 | 规则 | 问题 | 修复建议 |
|---|------|------|------|---------|
| 1 | docs/prd.md | 版本格式 | "版本:待定"不符合 v数字.数字 格式 | 改为 "版本:v0.1" |
| 2 | docs/prd.md | 字段非空 | "产品名称"为空 | 填写实际产品名称 |

### 🟡 警告(建议修复,不阻断)
| # | 文件 | 规则 | 问题 | 修复建议 |
|---|------|------|------|---------|
| 1 | docs/prd.md | 边界声明 | "明确不做"为空,缺少边界 | 至少写一条"本版不做什么" |

### 🔵 提示(最佳实践建议)
| # | 文件 | 建议 |
|---|------|------|
| 1 | docs/项目说明书.md | 「目录结构」建议填写,方便后续维护 |

### 流程审计结果(这一版新增)
| 检查 | 状态 |
|------|------|
| 阶段按序推进 | ✅ 一致 |
| 无假完成(出口门真过) | ❌ S1 标完成但 prd 缺验收 |
| 跳步均已留痕 | 🟡 S1.5 跳过未写理由 |
| 账本与产出物吻合 | ✅ 一致 |

### 项目进度卡审计结果
| 检查 | 状态 |
|------|------|
| 总表存在且非空 | ✅ 通过 |
| 总表只放索引级信息 | ✅ 通过 |
| 模块卡按需下钻 | 🟡 DU02 指向的模块卡不存在 |
| 任务 closeout 三处同步 | ✅ 通过 |
| 状态枚举合规 | ✅ 通过 |
| 下一轮 AI 最小上下文可用 | ✅ 通过 |

### 质量门结果
| 门 | 状态 |
|----|------|
| QG-001 文件存在 | ✅ 通过 |
| QG-002 无占位符 | ❌ 未通过 |
| QG-003 无敏感信息 | ✅ 通过 |
| QG-004 版本号 | ❌ 未通过 |
| QG-005 验收可验证 | ✅ 通过 |
| QG-006 语言检查 | ✅ 通过 |

### 总结
- 错误 X 项,警告 Y 项,提示 Z 项
- 质量门:A/B 通过
- 判定:❌ 不合格,需要修复后重新质检

如果全部通过:

## 质检报告:全部通过 ✅

所有产出物符合治理规范:
- 文件齐全、章节完整、字段合规
- 质量门 6/6 通过
- 跨 Skill 一致性检查通过

可以交付。

修复与重试

  1. 首次不通过:输出违规清单 + 每条的具体修复指令(用户可直接复制给 AI)
  2. 让用户把违规清单发给执行原任务的 AI,说"请按这些修复建议修正产出物"
  3. 修正后再次质检:只重新检查上次未通过的项
  4. 重试上限:最多 3 次。3 次后仍有错误未修复,标记为"⚠️ 需人工介入",列出剩余问题

使用方式

  • Claude Code:在 Skill 产出后直接说"帮我质检"或"检查一下产出"
  • ChatGPT/Codex:复制本文件内容,粘贴后说"请按上面的方法检查 docs/ 下的文件"
  • 自动化:配合 .claude/settings.json 中的 hooks,每次写入关键文件后自动提醒

速查:各 Skill 的检查要点

Skill关键检查项
vibe-coding-prdprd.md 必须章节齐全 / 版本号格式 / 产品名称非空 / 验收标准可验证
vibe-coding-requirements需求基准描述 80-300 字 / 覆盖四要素
vibe-coding-architecture选定技术栈含"选定"和"否决"
vibe-coding-production安全清单逐条确认 / 验收清单是 checkbox 格式
vibe-coding-survival以后再说清单存在
(流程)进度账本阶段按序、无假完成、跳步留痕、账本与产出物吻合
(开发期上下文)项目进度卡默认只读 index、模块按需下钻、任务 closeout 三处同步、状态枚举合规

与 CLAUDE.md 的关系

  • CLAUDE.md 是事前预防——在 AI 开始工作前就告诉它规则
  • 本 Skill 是事后校验——在 AI 声称完成后逐条验证
  • 两者都从 harness.json 派生,保持规则一致
  • 如果质检发现 CLAUDE.md 的规则没被遵守,修复产出 + 提醒用户下次开新对话时 CLAUDE.md 会生效

© Junliu1066, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/vibe-coding-harness of Junliu1066/vibe-coding-kit.

Open the folder on GitHubat commit cb387af

Compare with similar skills

Vibe Coding Harness next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Vibe Coding Harness compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Vibe Coding Harness this skillJunliu1066/vibe-coding-kit109—~1.3kAutomated safety check: PassMIT
CCPM Project Managementautomazeio/ccpm8.4k—~1.1kAutomated safety check: PassMIT
Ralph Tui Create Beadssubsy/ralph-tui2.5k1 repos~2.6kAutomated safety check: PassMIT
Trellis Brainstormanjiemo/SunnyBeach1787 repos~4kAutomated safety check: PassApache-2.0
Ralph Tui Create Beads Rustsubsy/ralph-tui2.5k1 repos~2.8kAutomated safety check: PassMIT
Ralph Tui Create JSONsubsy/ralph-tui2.5k1 repos~2.6kAutomated safety check: PassMIT

Similar skills

  • Runs a spec-driven workflow from PRD to epic to GitHub issues to parallel agents, with status, standup and blocked-work reports from bundled scripts.

    8.4k GitHub stars~1.1k tokensUpdated 6 mo ago
    Product & Project ManagementAuto-check passed
  • Ralph Tui Create Beads

    subsy/ralph-tui

    Convert PRDs to beads for ralph-tui execution. An agent skill from subsy/ralph-tui.

    2.5k GitHub starsUsed in 1 repo~2.6k tokens
    Product & Project ManagementAuto-check passed
  • Trellis Brainstorm

    anjiemo/SunnyBeach

    Guides collaborative requirements discovery before implementation.

    178 GitHub starsUsed in 7 repos~4k tokens
    Product & Project ManagementAuto-check passed
  • Convert PRDs to beads for ralph-tui execution using beads-rust (br CLI).

    2.5k GitHub starsUsed in 1 repo~2.8k tokens
    Product & Project ManagementAuto-check passed
  • Ralph Tui Create JSON

    subsy/ralph-tui

    Convert PRDs to prd.json format for ralph-tui execution. An agent skill from subsy/ralph-tui.

    2.5k GitHub starsUsed in 1 repo~2.6k tokens
    Product & Project ManagementAuto-check passed
  • Prd Generator

    jamesrochabrun/skills

    Generate comprehensive Product Requirements Documents (PRDs) for product managers.

    216 GitHub starsUsed in 2 repos~3.8k tokens
    Product & Project ManagementAuto-check passed

More from Junliu1066/vibe-coding-kit

  • Vibe Coding Prd

    Junliu1066/vibe-coding-kit

    一个"访谈式"的需求 Agent:通过一问一答 + 在岔路口给建议,把一个说不清需求的人,带到一份 AI 能直接照着开工的 PRD。

    109 GitHub stars~2k tokensUpdated 3 mo ago
    Auto-check passed
  • Vibe Coding Production

    Junliu1066/vibe-coding-kit

    把验证过的 Vibe Coding demo,做成能长期运行、给别人用的正式系统. An agent skill from Junliu1066/vibe-coding-kit.

    109 GitHub stars~993 tokensUpdated 3 mo ago
    Auto-check passed
  • Vibe Coding Requirements

    Junliu1066/vibe-coding-kit

    Vibe Coding(用 AI 写代码)项目的第一步:把模糊的想法,变成 AI 能精准落地、不会跑偏的需求. An agent skill from Junliu1066/vibe-coding-kit.

    109 GitHub stars~1.9k tokensUpdated 3 mo ago
    Auto-check passed
  • Vibe Coding Survival

    Junliu1066/vibe-coding-kit

    Vibe Coding(用 AI 写代码)开发过程中的避坑与自救——真正翻车大多发生在"开工后". An agent skill from Junliu1066/vibe-coding-kit.

    109 GitHub stars~857 tokensUpdated 3 mo ago
    Auto-check passed
  • Vibe Coding Architecture

    Junliu1066/vibe-coding-kit

    教非技术背景的人看懂技术架构、做好技术选型——把"只会照搬话术的小白"练成"能看穿方案好坏的人". An agent skill from Junliu1066/vibe-coding-kit.

    109 GitHub stars~1.2k tokensUpdated 3 mo ago
    Auto-check passed

Questions about Vibe Coding Harness

What does Vibe Coding Harness do?

质检 Agent:检查 vibe-coding-kit 其他 Skill 的产出物是否符合治理规范. An agent skill from Junliu1066/vibe-coding-kit. Vibe Coding Harness is an agent skill from Junliu1066/vibe-coding-kit.

When should I use Vibe Coding Harness?

Vibe Coding Harness fits situations like: tasks that involve PRD writing.

How do I install Vibe Coding Harness in Claude Code?

Run `npx skills add Junliu1066/vibe-coding-kit --skill vibe-coding-harness -a claude-code`. Or copy the skill folder (skills/vibe-coding-harness in Junliu1066/vibe-coding-kit) into .claude/skills/vibe-coding-harness in your project. Claude Code loads it when a task matches its description.

How do I install Vibe Coding Harness in Codex?

Run `npx skills add Junliu1066/vibe-coding-kit --skill vibe-coding-harness -a codex`. Or copy the skill folder (skills/vibe-coding-harness in Junliu1066/vibe-coding-kit) into .agents/skills/vibe-coding-harness in your project. Codex loads it when a task matches its description.

Can I use Vibe Coding Harness in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Junliu1066/vibe-coding-kit --skill vibe-coding-harness -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/vibe-coding-harness, .gemini/skills/vibe-coding-harness, .github/skills/vibe-coding-harness and .opencode/skills/vibe-coding-harness in your project.

What does Vibe Coding Harness need to run?

SKILL.md names no scripts, command-line tools or credentials: Vibe Coding Harness is instructions for the agent only.

Does Vibe Coding Harness access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Vibe Coding Harness safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Vibe Coding Harness use?

Vibe Coding Harness is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Vibe Coding Harness use?

About 1.3k tokens (SKILL.md is roughly 5.2k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Vibe Coding Harness?

Skills that share tags, products or a category with Vibe Coding Harness: CCPM Project Management (automazeio/ccpm, 8.4k stars), Ralph Tui Create Beads (subsy/ralph-tui, 2.5k stars), Trellis Brainstorm (anjiemo/SunnyBeach, 178 stars) and Ralph Tui Create Beads Rust (subsy/ralph-tui, 2.5k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Vibe Coding Harness?

Junliu1066 (a GitHub user) maintains it in Junliu1066/vibe-coding-kit, which has 109 GitHub stars. The repository holds 6 skills in this directory. The repository was last updated on June 30, 2026.

Source: Junliu1066/vibe-coding-kit on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.