Token导航 LogoToken导航

Agent Skill · 研究检索

writing-specs-designs(编写规格设计)

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-specs-designs 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,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 45,111。

总安装 870周安装 37GitHub Stars 3下载 305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-specs-designs(编写规格设计)
来源仓库:https://github.com/oldwinter/skills
仓库路径:skills/writing-specs-designs
安装命令:
npx skills add https://github.com/oldwinter/skills --skill writing-specs-designs
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/oldwinter/skills --skill writing-specs-designs

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。

用于辅助文档、README 和内容稿件的整理与改写。适合提炼结构、补齐章节、统一术语或检查链接有效性。通过 GitHub 安装并使用 npx skills add 命令集成到文档编写流程。应保留项目已有事实,避免把未确认信息写成确定结论。

writing-specs-designs 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/writing-specs-designs

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/oldwinter/skills --skill writing-specs-designs;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-specs-designs 怎么安装?

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

writing-specs-designs 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-specs-designs 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-specs-designs 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-specs-designs 用的人多吗?

公开数据是累计安装 870 次,近一周安装 37 次,下载 305 次,GitHub 3 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-specs-designs 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。
  • 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

  • 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-specs-designs

编写规格设计 · 本页

8703需要联网未明确通过

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

Writing Specs & Designs

Scope

Covers

  • Producing a Spec & Design Doc Pack that helps product, design, and engineering quickly reach shared clarity
  • Converting “insights” into concrete artifacts: a low‑fidelity diagram, flows/states, prototype plan, and testable acceptance criteria
  • Mobile-specific tap economy (when relevant): explicitly optimizing taps and interaction cost

When to use

  • “Write a spec / feature spec / technical spec for this feature.”
  • “Turn these notes into a design doc that engineering can build from.”
  • “Create a shaping-style doc with a diagram of the moving pieces.”
  • “We need a prototype plan to evaluate the feel of this interaction.”

When NOT to use

  • You’re still validating *what problem to solve* (do discovery / problem definition first).
  • You need a high-level strategy/vision document (do product vision / north star work first).
  • You need a deep engineering architecture design (APIs, schemas, scaling, reliability) rather than product interaction clarity.
  • You only need pixel-perfect UI mocks (this produces specs and structured guidance, not final visuals).

Inputs

Minimum required

  • Feature/change: 1–2 sentence description
  • Target users + primary use case (happy path)
  • Platform(s): web / iOS / Android / desktop (call out mobile explicitly)
  • Goals + non-goals (v1 boundaries)
  • Constraints: timeline, dependencies, policy/legal/privacy, technical constraints
  • Success metrics (1–3) + guardrails (2–5)

Missing-info strategy

  • Ask up to 5 questions from references/INTAKE.md, then proceed.
  • If key info remains missing, state assumptions explicitly and provide 2–3 options (scope/flows/prototype approach).

Outputs (deliverables)

Produce a Spec & Design Doc Pack in Markdown (in-chat; or as files if the user requests):

  1. Context snapshot (problem, why now, goals/non-goals, constraints, stakeholders)
  2. Low‑fidelity diagram of the moving pieces (≤10) + key decisions
  3. User flows + states (happy path + top edge cases) with clear entry/exit points
  4. Prototype brief (what to prototype, fidelity, timebox, data realism, success criteria)
  5. Requirements + acceptance criteria (testable; must/should/could)
  6. Measurement plan (metrics → data/events → owner → cadence)
  7. Risks / Open questions / Next steps (always included)

Templates: references/TEMPLATES.md

Workflow (8 steps)

1) Choose the smallest artifact set that unblocks the team

  • Inputs: User request + constraints.
  • Actions: Decide whether to deliver: (a) Spec & Design Doc only, or (b) Spec & Design Doc + Prototype Brief emphasis (feel-critical).
  • Outputs: Artifact selection + rationale.
  • Checks: Artifacts match the decision being made (alignment vs build execution vs interaction feel).

2) Intake + context snapshot

  • Inputs: references/INTAKE.md.
  • Actions: Ask up to 5 questions; confirm audience/DRI, platform(s), constraints, and success metrics/guardrails.
  • Outputs: Context snapshot section.
  • Checks: You can state the problem, user, and success measure in 1–2 sentences.

3) Lock scope boundaries (and define “tap budget” if mobile)

  • Inputs: Context snapshot.
  • Actions: Write goals, non-goals, out-of-scope, assumptions, dependencies. For mobile flows, define a “tap budget” (max taps to value) and where attention is most fragile.
  • Outputs: Scope boundaries + (optional) tap budget.
  • Checks: “What we are NOT doing” is explicit; mobile value path has a clear tap target.

4) Draft the low‑fidelity diagram (the “shaping” output)

  • Inputs: Scope + constraints.
  • Actions: Produce a diagram that shows the moving pieces (≤10), data/hand-offs, and where UX decisions matter. Avoid pixel-level UI; prefer clarity of structure.
  • Outputs: Low‑fidelity diagram + annotated decisions.
  • Checks: A partner can say “I know what to build” without a meeting.

5) Specify user flows + states (make edge cases concrete)

  • Inputs: Diagram + use case(s).
  • Actions: Document entry points, happy path, and top edge cases; include error/empty/loading states, permissions, and copy notes where ambiguity causes bugs.
  • Outputs: Flow + state table(s).
  • Checks: Another person can role-play the flow; each edge case has an intended outcome.

6) Write the prototype brief (only where “feel” matters)

  • Inputs: Flows/states + unknowns.
  • Actions: Identify the 1–3 riskiest interaction uncertainties; specify prototype type (low-fi, hi-fi, or in-code), timebox, realism (use real data where possible), and what “good” feels like.
  • Outputs: Prototype brief.
  • Checks: Prototype plan is testable (scenarios + success criteria), and explicitly disposable if using throwaway code.

7) Convert into testable requirements + acceptance criteria

  • Inputs: Flows/states + prototype learnings (if any).
  • Actions: Write requirements with acceptance criteria (must/should/could). Include non-functional needs (accessibility, performance, privacy) and instrumentation needs.
  • Outputs: Requirements section.
  • Checks: Engineering/QA can derive test cases without re-interpreting intent.

8) Quality gate + finalize for circulation

  • Inputs: Full draft pack.
  • Actions: Run references/CHECKLISTS.md and score with references/RUBRIC.md. Add Risks/Open questions/Next steps and clearly mark decisions vs assumptions.
  • Outputs: Final Spec & Design Doc Pack (shareable as-is).
  • Checks: Owners, metrics, and open questions are explicit; the diagram + flows remove the biggest ambiguities.

Quality gate (required)

Examples

Example 1 (mobile flow): “Write a spec + design doc for a new ‘invite friends’ flow in our iOS app. Goal: increase successful invites; minimize taps to first value.” Expected: tap budget + flow/state tables, a low-fi diagram of screens/transitions, and a prototype brief for the critical interaction.

Example 2 (B2B web feature): “Create a design doc/spec for ‘bulk edit roles’ for admins. Include edge cases and acceptance criteria.” Expected: permissions-focused flows/states, requirements with acceptance criteria, and a measurement plan.

Boundary example: “Write a spec to ‘improve engagement’ (no product context, no user, no success metric).” Response: ask the minimum intake questions; if still missing, provide 2–3 scoped options + assumptions and recommend upstream discovery before committing to a spec.

平台分布

Codex

33.2%
按下载量换算101

Claude

28.27%
按下载量换算86

Cursor

18.78%
按下载量换算57

Gemini CLI

9.95%
按下载量换算30

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

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。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