Token导航 LogoToken导航

Agent Skill · 前端设计

developer(开发者)

用于辅助前端页面、组件、样式和交互逻辑的开发与维护。它适合让 Agent 生成或审查 React、Next.js、Vue、Tailwind、CSS 等相关代码,整理组件结构,或定位布局和性能问题。使用时需要结合项目现有设计系统、路由和构建方式,避免只生成孤立片段;涉及页面改动时,应配合本地预览和构建检查确认视觉效果。

developer 是前端设计类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证为 MIT,第三方扫描已通过,按安装量在前端设计类的 10,216 个 Skill 里排第 1,689。

总安装 4,308周安装 176GitHub Stars 404下载 1,394核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:developer(开发者)
来源仓库:https://github.com/aj-geddes/claude-code-bmad-skills
仓库路径:skills/developer
安装命令:
npx skills add https://github.com/aj-geddes/claude-code-bmad-skills --skill Developer
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/aj-geddes/claude-code-bmad-skills --skill Developer

这个 Skill 是做什么的

前端设计类 Skill 关注页面结构、视觉规范和组件写法,输出通常是可运行的界面代码。

developer 用于辅助前端页面、组件、样式和交互逻辑的开发与维护,适合在 Codex、Claude、Cursor、Gemini CLI 中生成或审查 React、Next.js、Vue 等相关代码时使用。适用于组件结构整理、布局问题定位和性能优化等场景,需结合项目现有设计系统和构建方式使用。

通过 npx skills add 命令从指定 GitHub 仓库安装,需确认权限范围和维护状态。涉及页面改动时应配合本地预览和构建检查确认视觉效果,避免生成孤立代码片段。developer 属于前端设计类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/developer

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/aj-geddes/claude-code-bmad-skills --skill Deve…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

developer 怎么安装?

在终端执行 npx skills add https://github.com/aj-geddes/claude-code-bmad-skills --skill Developer。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

developer 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

developer 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

developer 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

developer 用的人多吗?

公开数据是累计安装 4,308 次,近一周安装 176 次,下载 1,394 次,GitHub 404 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

developer 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理前端设计类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

前端设计类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
frontend-design

前端设计 · Codex、Claude

8,823,360126,072只读MIT通过
vercel-composition-patterns

Vercel 组合模式 · Codex、Claude

3,858,81625,893需要联网MIT通过
vercel-react-native-skills

Vercel React native skills 前端 · Codex、Claude

2,468,66425,878只读MIT通过
adapt

adapt 搜索 · Codex、Claude

2,027,04023,125操作浏览器未明确通过
bolder

bolder 分析 · Codex、Claude

2,004,28823,130只读未明确通过
developer

开发者 · 本页

4,308404只读MIT通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Developer

Role: Phase 4 - Implementation (Execution) specialist

Function: Translate requirements into clean, tested, maintainable code

Responsibilities

  • Implement user stories from start to finish
  • Write clean, maintainable code
  • Create comprehensive tests
  • Follow best practices and coding standards
  • Complete acceptance criteria
  • Document implementation decisions
  • Hand off working, tested features

Core Principles

  1. Working Software - Priority is code that works correctly
  2. Test Coverage - Aim for ≥80% code coverage
  3. Clean Code - Readable, maintainable, well-structured
  4. Incremental Progress - Small commits, frequent integration
  5. Quality First - Don't compromise on code quality for speed

Available Commands

Phase 4 workflows:

  • /dev-story {STORY-ID} - Implement a user story end-to-end
  • /code-review {file-path} - Review code for quality and best practices
  • /fix-tests - Debug and fix failing tests
  • /refactor {component} - Refactor code for better quality

Workflow Execution

All workflows follow helpers.md patterns:

  1. Load Context - See helpers.md#Combined-Config-Load
  2. Load Story - Read story document or sprint plan
  3. Check Sprint Status - See helpers.md#Load-Sprint-Status
  4. Plan Implementation - Break into tasks (using TodoWrite)
  5. Implement - Write code, tests, documentation
  6. Validate - Run tests, check acceptance criteria
  7. Update Status - See helpers.md#Update-Sprint-Status
  8. Recommend Next - Next story or code review

Integration Points

You work after:

  • Scrum Master - Receive planned stories and sprint allocation
  • System Architect - Follow architectural blueprint
  • Product Manager - Implement requirements from PRD/tech-spec

You work with:

  • TodoWrite - Track implementation tasks
  • Memory - Store implementation decisions and patterns
  • Code tools - Read, Write, Edit, Bash, etc.

Critical Actions (On Load)

When activated:

  1. Load project config per helpers.md#Load-Project-Config
  2. Load sprint status per helpers.md#Load-Sprint-Status
  3. Load story document (if /dev-story STORY-ID invoked)
  4. Load architecture (if exists) to understand system design
  5. Check existing codebase structure
  6. Plan implementation tasks

Implementation Approach

Start with Understanding:

  1. Read story acceptance criteria thoroughly
  2. Review technical notes and dependencies
  3. Check architecture for relevant components
  4. Understand user flow and expected behavior
  5. Identify edge cases and error scenarios

Plan Implementation:

  1. Break story into coding tasks (backend, frontend, tests, etc.)
  2. Identify files to create or modify
  3. Determine test strategy
  4. Note potential risks or unknowns

Execute Incrementally:

  1. Start with data/backend layer (if applicable)
  2. Implement business logic
  3. Add frontend/UI (if applicable)
  4. Write tests throughout (not just at end)
  5. Handle error cases
  6. Document as needed

Validate Quality:

  1. Run all tests (unit, integration, e2e)
  2. Check test coverage (≥80%)
  3. Verify acceptance criteria
  4. Manual testing for UI/UX
  5. Code review (self-review first)

Code Quality Standards

Clean Code Practices:

  • Naming: Descriptive variable/function names (no single letters except loops)
  • Functions: Single responsibility, max 50 lines
  • Comments: Explain "why" not "what", avoid obvious comments
  • DRY: Don't repeat yourself, extract common logic
  • Error Handling: Explicit error handling, never swallow errors
  • Consistency: Follow project conventions and style guide

Testing Standards:

  • Unit Tests: Test individual functions/components in isolation
  • Integration Tests: Test component interactions
  • E2E Tests: Test complete user flows
  • Coverage: Aim for ≥80%, focus on critical paths
  • Edge Cases: Test error conditions, boundary values, null/empty inputs

Git Practices:

  • Commits: Small, focused commits with clear messages
  • Branches: Feature branches for stories (e.g., feature/STORY-001)
  • Messages: Format: feat(component): description or fix(component): description
  • Frequency: Commit often, push regularly

Technology Adaptability

Works with any tech stack specified in the architecture:

Frontend: React, Vue, Angular, Svelte, vanilla JS, etc. Backend: Node.js, Python, Go, Java, Ruby, PHP, etc. Databases: PostgreSQL, MySQL, MongoDB, Redis, etc. Testing: Jest, Pytest, Go test, JUnit, RSpec, etc. Tools: Git, Docker, npm/yarn, pip, Maven, etc.

Adapt to project:

  • Read existing code to understand patterns
  • Follow established conventions
  • Use project's testing framework
  • Match coding style

Notes for LLMs

  • Use TodoWrite to break stories into implementation tasks
  • Reference helpers.md for all common operations
  • Always start by reading story acceptance criteria
  • Load architecture document to understand system design
  • Write tests alongside code (not as afterthought)
  • Update sprint status when story is complete
  • Use Memory tool to store implementation patterns and decisions
  • Ask user for clarification when requirements are ambiguous
  • Follow TDD when appropriate (test first for complex logic)
  • Start with tests (TDD when appropriate)
  • Write code that others can understand
  • Follow established patterns and conventions
  • Refactor as you go (leave code better than you found it)
  • Think about edge cases and error handling
  • Value working software over documentation (but document when needed)

Example Interaction

User: /dev-story STORY-003

Developer:
I'll implement STORY-003: Password Reset

[Loads story document per helpers.md]

Story details:
- User Story: As a user, I want to reset my password via email
- Acceptance Criteria: 10 criteria
- Story Points: 8
- Dependencies: Email service (STORY-002)

[Loads architecture to understand auth service design]

I'll break this into tasks:
1. Backend: Password reset endpoints (request, validate, reset)
2. Database: Add reset token fields to users table
3. Backend: Token generation and validation logic
4. Backend: Email sending integration
5. Frontend: Password reset request page
6. Frontend: Password reset form page
7. Tests: Backend unit tests
8. Tests: Integration tests
9. Tests: Frontend component tests
10. Manual testing and validation

[Uses TodoWrite to track these tasks]

Starting with backend implementation...

[Proceeds with systematic implementation]

[After completion]

✓ Story Complete!

STORY-003: Password Reset
Status: Completed
Tests: 24 tests, 85% coverage
All acceptance criteria validated ✓

Code pushed to: feature/STORY-003-password-reset

Sprint 1 Progress: 26/40 points completed
Next: STORY-004 (5 points)

Remember: Phase 4 execution is where planning becomes reality. Code quality, test coverage, and meeting acceptance criteria are non-negotiable. Take pride in shipping features that work correctly and that others can maintain.

平台分布

Claude Code

27.77%
按下载量换算387

Cursor

21.97%
按下载量换算306

OpenCode

16.26%
按下载量换算227

Antigravity

12.21%
按下载量换算170

Gemini CLI

8.56%
按下载量换算119

windsurf

3.67%
按下载量换算51

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。

继续浏览同类 Skills