能力标签
💬
Prompt模板

克劳德代码工作流

基于 JavaScript · 专业级提示词模板,解锁 AI 的真实潜力
英文名:claude-code-workflows
⭐ 628 Stars 🍴 100 Forks 💻 JavaScript 📄 MIT 🏷 AI 8.2分
8.2AI 综合评分
promptclaude-codejavascriptai-agents
✦ AI Skill Hub 推荐

经 AI Skill Hub 精选评估,克劳德代码工作流 获评「强烈推荐」。这款Prompt模板在功能完整性、社区活跃度和易用性方面表现出色,AI 评分 8.2 分,适合有一定技术背景的用户使用。

📚 深度解析

克劳德代码工作流 是经过精心设计和实践验证的专业 Prompt 模板。Prompt 工程(Prompt Engineering)是充分发挥 Claude、ChatGPT 等大型语言模型潜力的关键技能,而一套经过优化的 Prompt 模板可以将 AI 输出质量提升数倍。

优质 Prompt 模板的核心价值在于其结构化设计:明确的角色设定、精确的任务描述、具体的输出格式要求和必要的边界条件,这些要素共同构成了一个能够持续产出高质量结果的 Prompt 框架。克劳德代码工作流 提供的模板经过反复迭代和用户验证,能够有效减少 AI 的"幻觉"(Hallucination)和输出不稳定问题。

无论你使用 Claude 3.5 Sonnet、GPT-4、Gemini 还是国内的文心一言、智谱 AI,优质的 Prompt 设计都能跨模型复用。AI Skill Hub 建议将本模板保存为个人 Prompt 库的标准组件,根据具体场景调整参数后反复使用,形成自己的 AI 提效工作流。

📋 工具概览

克劳德代码工作流 是经过精心设计和反复验证的专业 Prompt 模板集合。这些 Prompt 框架能够有效激活 Claude、ChatGPT 等大型语言模型的深层能力,让 AI 生成更准确、更有价值的输出结果。无需任何安装,直接复制模板内容到 AI 对话框即可使用。

GitHub Stars
⭐ 628
开发语言
JavaScript
支持平台
Windows / macOS / Linux
维护状态
正常维护,社区驱动
开源协议
MIT
AI 综合评分
8.2 分
工具类型
Prompt模板
Forks
100

📖 中文文档

以下内容由 AI Skill Hub 根据项目信息自动整理,如需查看完整原始文档请访问底部「原始来源」。

克劳德代码工作流 是经过精心设计和反复验证的专业 Prompt 模板集合。这些 Prompt 框架能够有效激活 Claude、ChatGPT 等大型语言模型的深层能力,让 AI 生成更准确、更有价值的输出结果。无需任何安装,直接复制模板内容到 AI 对话框即可使用。

📌 核心特色
  • 精心设计的 Prompt 框架,快速激活 AI 的深层能力
  • 支持参数化替换,灵活适配多种业务场景
  • 经过反复验证的指令结构,显著提升 AI 输出质量和一致性
  • 适用于 Claude、ChatGPT 等主流大语言模型
  • 可作为团队标准 Prompt 模板复用和二次开发
🎯 主要使用场景
  • 快速生成高质量的专业文案、分析报告或结构化内容
  • 利用 Prompt 框架引导 AI 解决特定领域的复杂问题
  • 在不同 AI 工具间复用经过验证的提示词模板
以下安装命令基于项目开发语言和类型自动生成,实际以官方 README 为准。
安装命令
# Prompt 无需安装,直接复制使用
# 支持:Claude / ChatGPT / Gemini / 通义千问 等主流模型

# 使用步骤
# 1. 复制 Prompt 模板内容
# 2. 粘贴到 AI 对话框
# 3. 替换 [占位符] 为实际内容
# 4. 发送后获取结构化输出

# 获取原始文件
git clone https://github.com/shinpr/claude-code-workflows
📋 安装步骤说明
  1. 复制本工具的 Prompt 模板内容
  2. 打开 Claude、ChatGPT 或其他 AI 对话工具
  3. 将 Prompt 粘贴到对话框开头
  4. 根据实际需求替换 [占位符] 中的内容
  5. 发送后 AI 将按照模板格式执行,获得结构化输出
以下用法示例由 AI Skill Hub 整理,涵盖最常见的使用场景。
常用命令 / 代码示例
# 粘贴到 Claude/ChatGPT 使用
# 示例 Prompt 结构:

你是一位 [角色],擅长 [领域]。
请根据以下要求完成任务:

任务背景:[描述背景]
具体要求:[详细说明]
输出格式:[期望格式]

# 将 [] 内内容替换为实际需求
以下配置示例基于典型使用场景生成,具体参数请参照官方文档调整。
配置示例
# claude-code-workflows 配置说明
# 查看配置选项
claude-code-workflows --config-example > config.yml

# 常见配置项
# output_dir: ./output
# log_level: info
# workers: 4

# 环境变量(覆盖配置文件)
export CLAUDE_CODE_WORKFLOWS_CONFIG="/path/to/config.yml"
📑 README 深度解析 真实文档 完整度 70/100 含工作流图 查看 GitHub 原文 →
以下内容由系统直接从 GitHub README 解析整理,保留代码块、表格与列表结构。

Backend Feature Development

```bash /recipe-implement "Add user authentication with JWT"

Frontend Feature Development

```bash /recipe-front-design "Build a user profile dashboard"

2. PRD covers the entire feature (single doc for all layers)

6. Work plan composes vertical feature slices

2. Generates PRD for each feature unit

1. requirement-analyzer determines scale

1. requirement-analyzer determines scale

Common setup

```bash

Install and start building

```bash

Install skills-only plugin

/plugin install dev-skills@claude-code-workflows


Skills auto-load when relevant. `coding-principles` activates during implementation, `testing-principles` during test writing, etc.

**Switching between plugins:**
bash

fixes lint and build errors, and commits per task.

```

Why UI Spec exists. Prototypes capture the visual surface, but rarely the loading, error, empty, and partial states that appear during integration.

For example, two components might each handle their own loading state cleanly while the dashboard that combines them has no defined behavior when one is still loading and the other has errored. UI Spec captures these state-x-display matrices and traces them into the Design Doc and test skeletons, so integration breakage is caught early.

⚡ Quick Start

Optional add-ons

These optional plugins cover adjacent stages in the workflow:

  • claude-code-discover: turns feature ideas into evidence-backed PRDs.
  • metronome: detects shortcut-taking behavior and nudges Claude to proceed step by step.
  • linear-prism: turns requirements into structured Linear tasks. Validates before decomposing, so downstream design starts clean.
  • pr-review: reviews GitHub PRs with Claude Code or Codex reviewers on deterministic PR snapshots, repo-specific quality criteria, and severity-gated posting with approval.
/plugin install discover@claude-code-workflows
/plugin install metronome@claude-code-workflows
/plugin install linear-prism@claude-code-workflows
/plugin install pr-review@claude-code-workflows

4. Optional prototype code is placed under docs/ui-spec/assets/

3. UI Spec captures the screen structure (with optional prototype)

Fullstack option: generates both backend and frontend Design Docs per feature unit

```

If you're working with undocumented legacy code, these commands generate PRD and design docs to make it AI-friendly. For a quick walkthrough, see: How I Made Legacy Code AI-Friendly with Auto-Generated Docs

---

Claude Code Workflows 🚀

Claude Code GitHub Stars License: MIT PRs Welcome

End-to-end development workflows for Claude Code that produce code aligned with design docs and tests. Specialized agents handle requirements, design, implementation, and quality checks, so what you get is code that actually passes its tests and matches its design docs.

---

Which plugin should I start with?

Your workInstall
Backend, APIs, CLI tools, or general programmingdev-workflows
React / TypeScript frontenddev-workflows-frontend
Full-stack (backend + React)dev-workflows-fullstack (single plugin, replaces installing both)
You already have your own orchestration and just want the rulesdev-skills (skills only, see [Skills only](#skills-only))

Fullstack (single plugin)

/plugin install dev-workflows-fullstack@claude-code-workflows /reload-plugins /recipe-fullstack-implement "Add user authentication with JWT + login form"

dev-skills -> dev-workflows

/plugin uninstall dev-skills@claude-code-workflows /plugin install dev-workflows@claude-code-workflows

dev-workflows -> dev-skills

/plugin uninstall dev-workflows@claude-code-workflows /plugin install dev-skills@claude-code-workflows ```

<a id="warning-duplicate-skills"></a>

Warning: dev-skills and dev-workflows / dev-workflows-frontend / dev-workflows-fullstack share the same skills. Installing more than one of them makes skill descriptions appear multiple times in the system context. Claude Code limits skill descriptions to ~2% of the context window. Exceeding this limit causes skills to be silently ignored. For fullstack work, install only dev-workflows-fullstack (do not install it alongside dev-workflows or dev-workflows-frontend).

---

The Workflow

graph TB A[User Request] --> B[requirement-analyzer] B --> |"Large (6+ files)"| C[prd-creator] B --> |"Medium (3-5 files)"| CA[codebase-analyzer] B --> |"Small (1-2 files)"| E[Direct Implementation] C --> CA CA --> D[technical-designer] D --> CV[code-verifier] CV --> DR[document-reviewer] DR --> DS[design-sync] DS --> F[acceptance-test-generator] F --> G[work-planner] G --> H[task-decomposer] H --> I[task-executor] E --> I I --> J[quality-fixer] J --> K[Ready to Commit]

This figure shows how a request is routed by size (small / medium / large). Larger work goes through PRD, codebase analysis, and design before reaching implementation. Smaller work skips ahead.

The Diagnosis Workflow

graph LR P[Problem] --> INV[investigator] INV --> |Failure Points| VER[verifier] VER --> |Coverage Check| COV{Sufficient?} COV --> |Yes| SOL[solver] COV --> |No| INV SOL --> |Solutions + Steps| R[Report]

The diagnosis flow loops between investigator and verifier until path coverage is sufficient, then hands off to solver for tradeoff analysis.

The Reverse Engineering Workflow

graph TB subgraph Phase1[Phase 1: PRD Generation] CMD["/recipe-reverse-engineer"] --> SD[scope-discoverer unified] SD --> PRD[prd-creator] PRD --> CV1[code-verifier] CV1 --> DR1[document-reviewer] end subgraph Phase2[Phase 2: Design Doc Generation] TD[technical-designer] --> CV2[code-verifier] CV2 --> DR2[document-reviewer] DR2 --> DONE[Complete] end DR1 --> |"All PRDs Approved (reuse scope)"| TD

Reverse engineering runs in two phases. PRDs come first (one per discovered feature), then Design Docs reuse the scope without re-discovering it.

⚡ Workflow Recipes

All workflow entry points use the recipe- prefix to distinguish them from knowledge skills. Type /recipe- and use tab completion to see all available recipes.

Backend & General Development (dev-workflows)

RecipePurposeWhen to Use
/recipe-implementEnd-to-end feature developmentNew features, complete workflows
/recipe-taskExecute single task with precisionBug fixes, small changes
/recipe-designCreate design documentationArchitecture planning
/recipe-planGenerate work plan from designPlanning phase
/recipe-prepare-implementationVerify implementation readiness and resolve gapsPre-build check that the work plan is implementable
/recipe-buildExecute from existing task planResume implementation
/recipe-reviewVerify code against design docsPost-implementation check
/recipe-diagnoseInvestigate problems and derive solutionsBug investigation, root cause analysis
/recipe-reverse-engineerGenerate PRD/Design Docs from existing codeLegacy system documentation, codebase understanding
/recipe-add-integration-testsAdd integration/E2E tests to existing codeTest coverage for existing implementations
/recipe-update-docUpdate existing design documents with reviewSpec changes, review feedback, document maintenance

For fullstack recipes, install dev-workflows-fullstack and use /recipe-fullstack-implement or /recipe-fullstack-build.

Frontend Development (dev-workflows-frontend)

The frontend plugin adds React-specific agents (component architecture, Testing Library, TypeScript-first quality checks) and UI Spec generation from optional prototype code.

RecipePurposeWhen to Use
/recipe-front-designCreate UI Spec + frontend Design DocReact component architecture, UI Spec
/recipe-front-planGenerate frontend work planComponent breakdown planning
/recipe-front-buildExecute frontend task planResume React implementation
/recipe-front-adjustAdjust an already-implemented UI with MCP-driven verificationVisual tweaks, component refinements that need design-source comparison
/recipe-front-reviewVerify code against design docsPost-implementation check
/recipe-taskExecute single task with precisionComponent fixes, small updates
/recipe-diagnoseInvestigate problems and derive solutionsBug investigation, root cause analysis
/recipe-update-docUpdate existing design documents with reviewSpec changes, review feedback, document maintenance
Tip: Both plugins share /recipe-task, /recipe-diagnose, and /recipe-update-doc. /recipe-update-doc auto-detects the document's layer. If your project has frontend Design Docs, the frontend plugin is needed to update them. For reverse engineering, use /recipe-reverse-engineer with the fullstack option to generate both backend and frontend Design Docs in a single workflow.

---

Shared Agents (Available in Both Plugins)

These agents work the same way whether you're building a REST API or a React app:

AgentWhat It Does
**requirement-analyzer**Figures out how complex your task is and picks the right workflow
**codebase-analyzer**Analyzes existing codebase before design to produce focused guidance for technical-designer
**work-planner**Breaks down design docs into actionable tasks
**task-decomposer**Splits work into small, commit-ready chunks
**code-reviewer**Checks your code against design docs to make sure nothing is missing
**document-reviewer**Reviews single document quality, completeness, and rule compliance
**design-sync**Verifies consistency across multiple Design Docs and detects conflicts
**investigator**Maps execution paths, identifies failure points with causal chains for problem diagnosis
**verifier**Validates failure points, checks path coverage using Devil's Advocate method
**solver**Generates solutions with tradeoff analysis and implementation steps
**security-reviewer**Reviews implementation for security compliance after all tasks complete

Backend-Specific Agents (dev-workflows)

AgentWhat It Does
**prd-creator**Writes product requirement docs for complex features
**technical-designer**Plans architecture and tech stack decisions
**scope-discoverer**Discovers functional scope from codebase for reverse engineering
**code-verifier**Validates consistency between documentation and code implementation
**acceptance-test-generator**Creates E2E and integration test scaffolds from requirements
**integration-test-reviewer**Reviews integration/E2E tests for skeleton compliance and quality
**task-executor**Implements backend features with TDD
**quality-fixer**Runs tests, fixes type errors, handles linting (everything quality-related)
**rule-advisor**Picks the best coding rules for your current task

Frontend-Specific Agents (dev-workflows-frontend)

AgentWhat It Does
**prd-creator**Writes product requirement docs for complex features
**ui-spec-designer**Creates UI Specifications from PRD and optional prototype code
**ui-analyzer**Reads the project's external-resources file, fetches design / design-system / guideline sources via inherited MCP access, and analyzes existing UI code
**technical-designer-frontend**Plans React component architecture and state management
**task-executor-frontend**Implements React components with Testing Library
**code-verifier**Validates consistency between documentation and code implementation
**quality-fixer-frontend**Handles React-specific tests, TypeScript checks, and builds
**rule-advisor**Picks the best coding rules for your current task
**design-sync**Verifies consistency across multiple Design Docs and detects conflicts

---

🎯 Typical Workflows

5. UI Spec captures screen structure, components, and interactions

Implements components with Testing Library, handles TypeScript types,

Fullstack Workflow

```bash /recipe-fullstack-implement "Add user authentication with JWT + React login form"

9. Vertical slices commit early so integration is verified per phase

```

Install dev-workflows-fullstack for fullstack recipes. Tasks are routed based on filename patterns (*-backend-task-*, *-frontend-task-*). For reverse engineering existing fullstack codebases, use /recipe-reverse-engineer with the fullstack option.

UI Adjustment (Frontend Plugin)

```bash /recipe-front-adjust "Tighten Card spacing and align action buttons with the design source"

Quick Fixes (Both Plugins)

```bash /recipe-task "Fix validation error message"

Direct implementation with quality checks. Works the same in both plugins.

```

Problem Diagnosis (Both Plugins)

```bash /recipe-diagnose "API returns 500 error on user login"

🔌 Contributing External Plugins

This marketplace supports the full lifecycle of building products with AI: product quality, discovery, implementation control, and verification. If your plugin helps developers build better products with AI coding agents, we'd like to hear from you.

See CONTRIBUTING.md for submission guidelines and acceptance criteria.

---

🤔 FAQ

Q: Do I need to learn special commands?

A: Not really. For backend, just use /recipe-implement. For frontend, use /recipe-front-design. The plugins handle everything else automatically.

Q: What if there are errors?

A: The quality-fixer agents (one in each plugin) automatically fix most issues like test failures, type errors, and lint problems. If something cannot be auto-fixed, you'll get clear guidance on what needs attention.

Q: Is there a version for OpenAI Codex CLI?

A: Yes. codex-workflows provides the same end-to-end development workflows for Codex CLI. Same concept (specialized subagents for requirements, design, implementation, and quality checks), adapted for the Codex CLI environment.

Q: Should I commit the work plan and task files in docs/plans/?

A: No. Recipes treat docs/plans/ as ephemeral working state. Work plans, task files, prep tasks, review-fix tasks, and intermediate analysis files all live there during a recipe run and are cleaned up at the end. Add the following line to your project's .gitignore so working state stays out of git:

docs/plans/

PRDs, ADRs, UI Specs, and Design Docs live in their own directories (docs/prd/, docs/adr/, docs/ui-spec/, docs/design/) and are intended to be committed.

---

🎯 aiskill88 AI 点评 A 级 2026-07-12

高质量的开源Prompt模板,适合生产环境

⚡ 核心功能

👥 适合人群

内容创作者和自媒体人职场人士和学生ChatGPT / Claude 重度用户希望提升 AI 使用效率的普通用户

🎯 使用场景

  • 快速生成高质量的专业文案、分析报告或结构化内容
  • 利用 Prompt 框架引导 AI 解决特定领域的复杂问题
  • 在不同 AI 工具间复用经过验证的提示词模板

⚖️ 优点与不足

✅ 优点
  • +MIT 协议,可免费商用
  • +无需安装,立即可用
  • +适配所有主流 AI 工具
  • +经社区验证的最佳实践
⚠️ 不足
  • 效果依赖使用者对 Prompt 工程的熟悉程度
  • 不同模型和版本的响应效果可能存在差异
  • 复杂场景需结合实际需求二次调整
⚠️ 使用须知

AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。

建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。

📄 License 说明

✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。

🔗 相关工具推荐

🧩 你可能还需要
基于当前 Skill 的能力图谱,自动补全的工具组合

❓ 常见问题 FAQ

Not really. For backend, just use `/recipe-implement`. For frontend, use `/recipe-front-design`. The plugins handle everything else automatically.
💡 AI Skill Hub 点评

AI Skill Hub 点评:克劳德代码工作流 的核心功能完整,质量优秀。对于内容创作者和自媒体人来说,这是一个值得纳入个人工具库的选择。建议先在非生产环境试用,再逐步推广。

⬇️ 获取与下载
⬇ 下载源码 ZIP

✅ MIT 协议 · 可免费商用 · 直接从 aiskill88 服务器下载,无需跳转 GitHub

📚 深入学习 克劳德代码工作流
查看分步骤安装教程和完整使用指南,快速上手这款工具
🌐 原始信息
原始名称 claude-code-workflows
Topics promptclaude-codejavascriptai-agents
GitHub https://github.com/shinpr/claude-code-workflows
License MIT
语言 JavaScript
🔗 原始来源
🐙 GitHub 仓库  https://github.com/shinpr/claude-code-workflows

收录时间:2026-07-12 · 更新时间:2026-07-12 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。

📺 订阅 AI Skill Hub Daily Telegram 频道
每天 8 条精选 AI Skill、MCP、Agent 与自动化工具推送
加入频道 →