Token导航 LogoToken导航

Agent Skill · 研究检索

prd-writing(PRD 写作)

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。

prd-writing 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 39,984。

总安装 1,344周安装 56GitHub Stars 33下载 448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:prd-writing(PRD 写作)
来源仓库:https://github.com/asyrafhussin/agent-skills
仓库路径:skills/prd-writing
安装命令:
npx skills add https://github.com/asyrafhussin/agent-skills --skill prd-writing
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/asyrafhussin/agent-skills --skill prd-writing

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。

用于辅助产品需求文档(PRD)撰写,帮助 Agent 提炼结构、统一术语并整合零散材料为清晰可读的 Markdown 文档。它提供分步工作流程与 25 条规则,涵盖用户故事拆解、验收条件定义和技术约束说明。使用时需先探索项目代码基,确认已有模型和路由结构,避免脱离实际架构虚构功能点;对外文案需控制语气,不夸大能力。

输出结果仅为草稿建议,必须经过产品经理或技术负责人复核后方可进入开发阶段。prd-writing 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/prd-writing

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/asyrafhussin/agent-skills --skill prd-writing;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

prd-writing 怎么安装?

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

prd-writing 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。

prd-writing 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。

prd-writing 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。

prd-writing 用的人多吗?

公开数据是累计安装 1,344 次,近一周安装 56 次,下载 448 次,GitHub 33 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。

prd-writing 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
prd-writing

PRD 写作 · 本页

1,34433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

PRD Writing

Step-by-step workflow for writing clear, actionable Product Requirements Documents. Follow the 6-step process below, using the 25 rules across 7 categories as supporting knowledge.

Metadata

  • Version: 1.0.0
  • Rule Count: 25 rules across 7 categories
  • License: MIT

Workflow

Follow these steps in order. Skip a step only if the user has already provided that information. Do NOT start drafting before completing discovery.

Step 1: Assess Project State

Determine what you're working with:

  • Existing project — code exists → explore the codebase first (models, routes, controllers, auth, patterns). See disc-codebase-exploration.
  • Empty/greenfield project — no code yet → ask the user what they want to build first. Brainstorm together until the product idea and core features are clear, then ask about planned stack, architecture decisions, and constraints.

Step 2: Ask Clarifying Questions

Ask 3-5 targeted questions to fill knowledge gaps. Use lettered options (A/B/C) when there are clear choices, and open-ended questions when you need free-text answers. Focus on:

  1. Problem — What problem are we solving? Why now?
  2. Users — Who are the target users?
  3. Scope — What should it NOT do?
  4. Success — How do we know it's done?
  5. Constraints — Timeline, budget, team size?

Mark unknowns as TBD, not assumptions. See disc-clarifying-questions.

Step 3: Draft the PRD

Use the template from struct-prd-template and fill in all 12 sections:

  1. Executive Summary — 1 paragraph: problem, solution, impact
  2. Problem Statement — what, who, evidence, why now
  3. Goals & Success Metrics — 3-5 measurable KPIs
  4. User Personas — primary and secondary with roles/goals/pain points
  5. User Stories & Acceptance Criteria — "As a / I want / So that" with checkboxes
  6. Functional Requirements — numbered FR-1, FR-2, etc.
  7. Non-Functional Requirements — performance, security, accessibility with specific numbers
  8. Technical Specifications — data model, auth, routes, integrations
  9. Out of Scope — what we're NOT building and why
  10. Non-Goals — what we're NOT optimizing for
  11. Dependencies & Risks — blockers with owners and mitigations
  12. Open Questions — unresolved items with owners and due dates

Key rules to follow while drafting:

  • Start with the problem, not the solution (disc-problem-first)
  • Replace vague language with numbers (metric-no-vague-language)
  • Every requirement must be testable (quality-acceptance-criteria)
  • Use Given/When/Then for complex acceptance criteria

Step 4: Present for Review

Show the draft to the user. Ask specifically:

  • Does the problem statement match your understanding?
  • Are user stories missing any workflows?
  • Are the scope boundaries correct?
  • Any open questions I should resolve?

Step 5: Revise Based on Feedback

Incorporate feedback, resolve open questions, and update the PRD. Add a review history entry if the PRD will be shared with a team.

Step 6: Save the PRD

Save to docs/prd/{feature-name}.md using kebab-case naming. Include frontmatter:

---
title: Feature Name
status: draft
author: Author Name
created: YYYY-MM-DD
updated: YYYY-MM-DD
---

Rules Reference

The 25 rules below provide detailed guidance for each step. Read them when you need deeper context.

1. Discovery (CRITICAL)

  • disc-problem-first - Start with the problem, not the solution
  • disc-clarifying-questions - Ask clarifying questions before writing
  • disc-codebase-exploration - Explore the codebase before drafting
  • disc-stakeholder-alignment - Align with stakeholders on goals and constraints

2. Structure (CRITICAL)

  • struct-standard-sections - Use standardized PRD sections
  • struct-executive-summary - Write a concise executive summary
  • struct-prd-template - Provide a ready-to-use PRD template
  • struct-output-location - Save PRDs to a consistent file location
  • struct-single-source-of-truth - PRD is the definitive reference for a feature

3. Requirements (HIGH)

  • req-user-personas - Define target user personas
  • req-user-stories - Write user stories with acceptance criteria
  • req-functional - Define specific functional requirements
  • req-non-functional - Define non-functional requirements

4. Scope (HIGH)

  • scope-out-of-scope - Explicitly define what is out of scope
  • scope-non-goals - List non-goals to protect timeline
  • scope-dependencies - Identify dependencies and blockers

5. Metrics (HIGH)

  • metric-measurable-success - Define measurable success criteria
  • metric-no-vague-language - Replace vague terms with quantifiable benchmarks
  • metric-kpis - Define 3-5 key performance indicators

6. Technical (MEDIUM)

  • tech-data-model - Document the data model and relationships
  • tech-auth-model - Define authentication and authorization
  • tech-api-routes - Document API and route structure
  • tech-integration-points - Identify third-party integrations

7. Quality (MEDIUM)

  • quality-acceptance-criteria - Write testable acceptance criteria
  • quality-iterative-review - Iterate with feedback before finalizing

References

Full Compiled Document

For the complete guide with all rules expanded: AGENTS.md

平台分布

Codex

34.88%
按下载量换算156

Claude

32.94%
按下载量换算148

Cursor

17.99%
按下载量换算81

Gemini CLI

9.88%
按下载量换算44

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