需求成文 - 产品经理 AI 工作台

SKILLAI 工具
AI 工具 未标注权限 许可证不明确
安装这个技能

将业务想法、材料或已有文档整理为可编辑的 Domain Model 和 PRD,通过独立新会话检查需求,并以决策档案保留项目上下文。适用于需求澄清、业务建模、PRD 编写与修订、已有 PRD 检查,以及跨会话接续需求工作。

一句话选型结论

按资料字段整理

需求成文 - 产品经理 AI 工作台是AI 工具类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类AI 工具 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_1e4e8da7/pm-ai-copilot-requirements
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

AI 工具
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_1e4e8da7仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验2026-11-06上游最后核验时间 · 数据更新 2026-11-06
正文体量7.6 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
组合多个 AI 能力

把写作、检索、总结串成一条流程。

搭工作流

把固定流程做成可重复调用的步骤。

做提示词与评测

整理提示词并对比输出效果。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在AI 工具场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
需求成文 - 产品经理 AI 工作台当前 未标注权限 未标注 未标注 未标注 有
执行计划 只读 GitHub 仓库 151.5 万 Codex、Claude、Cursor 有
Playwright MCP 测试 浏览器控制 ClawHub 市场 136.5 万 OpenClaw 有
完成前的验证 未标注权限 GitHub 仓库 131.9 万 Codex、Claude、Cursor 有
子 Agent 驱动开发 只读 GitHub 仓库 131.2 万 Codex、Claude、Cursor 有
前端设计 执行命令 GitHub 仓库 125.6 万 Codex、Claude、Cursor 有

SKILL.md 全文

7.6 KB
模型实际读到的说明书原文,来自技能仓库

<!-- Adapted from PM Agentic Cookbook source-intake, domain-modeling, create-prd and check-prd; modified by PM Agentic Cookbook contributors. SPDX-License-Identifier: Apache-2.0. See LICENSE and NOTICE. -->

PM AI Copilot — Requirements

Deliver professional, editable requirements with traceable business meaning. Support text in conversation and readable Markdown/plain-text files. Use host capabilities for other formats only when actually available and checked. Do not promise Word/PDF conversion or treat unreadable extracts as evidence. This skill produces requirements and checks; prototype generation and production implementation are separate tasks.

Choose the task and role first

  • For an independent check, read review instructions and execute the reviewer role directly. Do not first load author notes or project memory. The caller supplies the target and approved comparison materials; missing context limits the check, not the source's legitimacy.
  • For new requirements, an increment or a resumed authoring task, read project contract. Locate the user-selected project and actual current materials. Ask only if its directory, effective version or authority is unclear.

Resolve roles: this skill defines generic requirement work; a Provider is a set of business rules or conventions explicitly selected by the user; the Consumer supplies the current request, project locations and host capabilities. Do not infer or load an industry Provider from vocabulary alone. State which supplied business rules apply, or that none were supplied. Do not require a repository-wide setup, a separate acceptance checklist or another installed package.

Authoring path

  1. Read and clarify. Use intake and modeling. Identify sources, confirmed facts, assumptions, conflicts and questions. Ask the smallest set of questions that changes the goal, scope, objects, relationships, states or rules. Preserve an oral brief in the project state so it can be referenced later. Do not force a material tier questionnaire or multiple intake reports.
  2. Build the Domain Model. Read the Domain Modeling workflow and use its five-step method and model template. Preserve entity/field specifications, ER and relationship details, state, calculations, rules and applicable temporal/event constraints. Name new business documents <scope>_domain_model.md and <scope>_prd.md, matching their full titles; follow existing mappings or an authorized rename. Link them to sources. Present the model and scope for user confirmation; stop before generating a new PRD until they are confirmed. A previous explicit confirmation remains valid for its unchanged scope. Synchronize that actual confirmation into the current document and state; do not ask for it again as a ritual.
  3. Write or revise the PRD. Read the derived Create PRD workflow, its product-typing and applicable chapter references; use the 14-chapter PRD template. Preserve applicable detailed A–I page rules, permission structure, Overview/Sub ownership and the original self-check. Include core user scenarios, observable expected results and exceptions inside the PRD. Preserve existing authoritative document locations and identifiers. Load diagram rules when a diagram is needed. Do not require a separate acceptance-standard file or ledger.
  4. Check as author, then hand off. Save the complete author self-check separately from the business document. Read review instructions to prepare a clean handoff into a new WorkBuddy conversation. Give paths and exact reviewed versions, approved source/model, relevant rules, instructions and a unique report location. Freeze the business inputs and local resources; extract a section-mapped review copy for mixed documents. Omit author assessments, expected defects and previous conclusions from both the request and first-stage inputs. Rechecks deliver old reports only after the current judgment is written. Ask the user to open that independent conversation; do not continue in the author's context and call it independent. The installed resource path is resolved from this skill, never from a hard-coded machine path.
  5. Handle real findings. Read the actual independent report, present its findings and disagreements accurately, and obtain any required business decision or edit authority. Fix only the authorized scope, preserve previous content as required by the project contract, and request a separate check of changed items and adjacent effects. A reported issue, an author fix and a reviewer recheck are distinct results.
  6. Close the session. Read back real deliverables and update the project state, pending questions, meaningful history and next action. Record actual choices and reasons or say no new business decision was made. Return artifact locations, confirmation states, checks and remaining limits. Do not claim approval or acceptance on the user's behalf.

Stop affected work

Stop and describe the needed input when the project location or effective authority is ambiguous, a key business rule is unsupported, sources conflict, confirmation is missing, a required reference cannot be read, or a proposed edit exceeds authorization. Unaffected work may continue when its boundary is clear. Never fill a gap with a plausible enterprise default and label it confirmed.

For an independent check, decline the independent label if you authored the target or inherited author conclusions. A new conversation with only the review contract is the supported handoff. Do not promise invisible process or memory isolation. If that handoff cannot be performed, return the request and mark the independent check unperformed.

Source and license

This skill is derived from the author's open-source project PM Agentic Cookbook. Copyright 2026 Sunnyshinnny776 and contributors. Its PRD creation and checking workflows include adaptations of 杨堃 (pmYangKun)'s create-prd-skill and check-prd-skill, through Cookbook. Preserve their authorship, methodology attribution and applicable contributor notices. The upstream notices, including the applicable @Scofy0123 contribution, remain in NOTICE; the Apache-2.0 text is in LICENSE. The adaptations do not imply upstream endorsement.

常见问题

5 问
答案由本站依据条目字段整理
需求成文 - 产品经理 AI 工作台怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_1e4e8da7/pm-ai-copilot-requirements。安装方式为未标注。装完刷新技能列表或重开客户端。

需求成文 - 产品经理 AI 工作台支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

需求成文 - 产品经理 AI 工作台权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

需求成文 - 产品经理 AI 工作台收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

需求成文 - 产品经理 AI 工作台用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 2026-11-06。