Token导航 LogoToken导航

Agent Skill · 研究检索

writing-design-docs(编写设计文档)

用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写。它适合让 Agent 提炼结构、补齐章节、统一术语、检查链接或把零散材料整理成可读文档。使用时应保留项目已有事实、命令和路径,不要把未确认的信息写成确定结论;涉及对外文案时,还需要控制语气,避免过度营销或夸大能力。

writing-design-docs 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 43,282。

总安装 998周安装 40GitHub Stars 1下载 323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
unknown

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:writing-design-docs(编写设计文档)
来源仓库:https://github.com/neekolas/claude-skills
仓库路径:skills/writing-design-docs
安装命令:
npx skills add https://github.com/neekolas/claude-skills --skill writing-design-docs
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。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/neekolas/claude-skills --skill writing-design-docs

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

2 句话 3句话 4句话 目标/非目标 每颗2-3颗子弹 每颗4-5颗子弹 6+ 具有不变量 背景 仅链接 链接+文件列表 链接+文件+影响 系统设计 1 段 接口+功能 图表 + 完整的 API 界面 图书馆 表或“无” 表+理由 表 + 替代方案 测试 3-5耳 8-12 耳 + 边缘情况 15+耳朵+综合优势 常见错误 错误 修复 事后才进行测试 将红/绿 TDD 与规范中定义的测试用例结合使用 目标模糊 添加数字:“将 p99 从 800ms 减少到 200ms” 缺少非目标 未说明的范围 = 假定的范围 实施即设计 契约和行为,而不是代码 没有上下文链接 链接催化剂——未来的读者需要“为什么” 示例 参见示例/ 对于分级示例: 小-cli-flag.md — 添加 --verbose 标志(最小但完整) 中型 API 端点.md — 具有身份验证和速率限制的 REST API 大型事件系统.md — 具有重试和 DLQ 的分布式事件管道 每周安装量 40 存储库 尼科拉斯/克劳德-技能 GitHub 之星 1 第一次看到 6 天前 安全审计 Gen Agent Trust Hub 通行证 套接字通行证 斯尼克警告

来源仓库内路径:skills/writing-design-docs

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/neekolas/claude-skills --skill writing-design-…;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

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

writing-design-docs 怎么安装?

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

writing-design-docs 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。

writing-design-docs 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

writing-design-docs 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

writing-design-docs 用的人多吗?

公开数据是累计安装 998 次,近一周安装 40 次,下载 323 次,GitHub 1 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。

writing-design-docs 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

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。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
writing-design-docs

编写设计文档 · 本页

9981需要联网未明确有提醒

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

Writing Design Docs

Produce a structured design document that captures what you're building, why, and how you'll verify it works. Output: docs/plans/YYYY-MM-DD-<topic>-design.md. This skill produces the artifact and stops.

Process

digraph design_doc {
    "Explore codebase & context" [shape=box];
    "Propose 2-3 approaches" [shape=box];
    "User selects approach" [shape=diamond];
    "Draft each section" [shape=box];
    "Write EARS requirements" [shape=box];
    "Present to user for review" [shape=box];
    "User approves?" [shape=diamond];
    "Write to docs/plans/" [shape=box];
    "Done" [shape=doublecircle];

    "Explore codebase & context" -> "Propose 2-3 approaches";
    "Propose 2-3 approaches" -> "User selects approach";
    "User selects approach" -> "Propose 2-3 approaches" [label="none fit"];
    "User selects approach" -> "Draft each section" [label="selected"];
    "Draft each section" -> "Write EARS requirements";
    "Write EARS requirements" -> "Present to user for review";
    "Present to user for review" -> "User approves?";
    "User approves?" -> "Draft each section" [label="revise"];
    "User approves?" -> "Write to docs/plans/" [label="yes"];
    "Write to docs/plans/" -> "Done";
}

Proposing Approaches

Before drafting, explore the solution space. Research the codebase and relevant libraries, then propose 2-3 approaches:

  • Lead with your recommendation and explain why
  • For each approach: architecture summary, pros, cons, effort estimate
  • Include at least one simpler/smaller and one more robust/extensible option
  • Let the user select before committing to a design direction

Design Doc Template

Six sections. Scale depth to complexity — see Scaling Guide.

1. Summary

2-4 sentences. What and why. A busy engineer should understand the project from this alone.

2. Project Goals & Non-Goals

Goals: Problem being solved. Invariants that must hold. Be specific — "p99 under 200ms", not "fast".

Non-Goals: Reasonable things explicitly out of scope. Not negated goals — things you're choosing not to address.

3. Context

  • Catalysts: GitHub Issues, Slack threads, or other triggers
  • Codebase: Existing folders, files, and design docs relevant to this work
  • External docs: URLs for third-party library documentation
  • References: Blog posts, RFCs, or source materials that informed the design
  • Impact area: Modules or directories that will be modified

4. System Design

  • Architecture overview: How components fit together. Diagram if helpful.
  • New or modified interfaces: Class/struct definitions, API boundaries. Shape, not implementation.
  • Key functions: Important functions with expected behavior. Contracts and invariants.
  • Alternatives considered: Why rejected approaches didn't make the cut.

5. Libraries & Utilities Required

External dependencies:

PackageVersionPurpose
name^x.y.zWhy needed

Internal modules:

ModulePathPurpose
namesrc/path/What it provides

Write "None" if no dependencies — don't omit the section.

6. Testing & Validation

This is the most important section. It should be the most detailed.

Acceptance Criteria

Use EARS format for every criterion. Each must be testable and unambiguous.

Edge Cases

Address relevant categories: concurrency/race conditions, dependency failures, error handling/recovery, boundary conditions, security considerations.

Verification Commands

Concrete commands to prove correctness. Include linting and formatting checks.

EARS Quick Reference

PatternTemplateExample
UbiquitousTHE SYSTEM SHALL [behavior]THE SYSTEM SHALL encrypt all data at rest
Event-drivenWHEN [event] THE SYSTEM SHALL [behavior]WHEN a request exceeds the rate limit THE SYSTEM SHALL return HTTP 429
State-drivenWHILE [state] THE SYSTEM SHALL [behavior]WHILE the circuit breaker is open THE SYSTEM SHALL return cached responses
OptionalWHERE [feature] THE SYSTEM SHALL [behavior]WHERE verbose logging is enabled THE SYSTEM SHALL log request bodies
UnwantedTHE SYSTEM SHALL NOT [behavior]THE SYSTEM SHALL NOT expose internal error details to clients
ComplexWHEN [a] AND [b] THE SYSTEM SHALL [behavior]WHEN the queue is full AND the message is high-priority THE SYSTEM SHALL evict the oldest low-priority message

Rules: Use SHALL, never "should"/"may". Each requirement independently testable. No vague terms — use measurable criteria.

Scaling Guide

SectionSmall (~150w)Medium (~400w)Large (~800w)
Summary2 sentences3 sentences4 sentences
Goals/Non-Goals2-3 bullets each4-5 bullets each6+ with invariants
ContextLinks onlyLinks + file listLinks + files + impact
System Design1 paragraphInterfaces + functionsDiagram + full API surface
LibrariesTable or "None"Table + rationaleTable + alternatives
Testing3-5 EARS8-12 EARS + edge cases15+ EARS + comprehensive edges

Common Mistakes

MistakeFix
Testing as afterthoughtUse red/green TDD with test cases defined in the spec
Vague goalsAdd numbers: "reduce p99 from 800ms to 200ms"
Missing non-goalsUnstated scope = assumed in scope
Implementation as designContracts and behavior, not code
No context linksLink the catalyst — future readers need the WHY

Examples

See examples/ for graduated examples:

  • small-cli-flag.md — Adding a --verbose flag (minimal but complete)
  • medium-api-endpoint.md — REST API with auth and rate limiting
  • large-event-system.md — Distributed event pipeline with retry and DLQ

平台分布

Codex

37.6%
按下载量换算121

Claude

29.82%
按下载量换算96

Cursor

17.59%
按下载量换算57

Gemini CLI

10.46%
按下载量换算34

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills