论文生成
- 作者仓库星标 38
- 许可证 MIT
- 作者更新于 实时读取
- 作者仓库 builder-skills
- 领域
- 工程开发
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 94 / 100 · 已通过审计
- 作者 / 版本 / 许可
- @kazdenc · MIT
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 即装即用
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明（默认跨平台）
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生，可能与作者实际意图存在差异。
需要注意： 未限定 allowed-tools，默认拥有全部工具权限。
---
name: prd
description: Write a product requirements document from context. Covers problem statement, goals, scope, user…
category: 工程开发
runtime: 无特殊运行时
---
# prd 输出预览
## PART A: 任务判断
- 适用问题：代码实现、重构、调试或代码审查。
- 输入要求：目标材料、限制条件、期望输出和验收方式。
- 证据边界：围绕“Step 1: Gather Context / Step 2: Write the PRD / Problem Statement”读取原文规则，不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断：确认你的需求是否属于代码实现、重构、调试或代码审查，并标出输入、限制和预期结果。
- **02** 执行计划：优先按“Step 1: Gather Context / Step 2: Write the PRD / Problem Statement”拆成步骤，说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果：给出可复制的命令、文件改动、检查清单或内容草稿，并说明如何继续迭代。
- **04** 风险边界：结合 读取文件、写入/修改文件、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件；主要在本地完成；通常不需要额外 API Key。
- 先小样例验证，再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 先确认触发方式
原文没有稳定的斜杠命令要求。安装验证后通常全局生效，直接在对话里点名这个 Skill 并描述任务即可。
给清楚输入和边界
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是：读取文件、写入/修改文件。
小样例验证后再放大
先用一个小任务确认它会围绕“Step 1: Gather Context / Step 2: Write the PRD / Problem Statement”工作；涉及文件或命令时，先看 diff、日志、预览或测试结果。
复核后再交付
检查最终产物是否包含明确结果、必要证据和下一步动作；如果输出泛泛而谈，就补充输入、边界和验收标准后重跑。
---
name: prd
description: Write a product requirements document from context. Covers problem statement, goals, scope, user…
category: 工程开发
source: kazdenc/builder-skills
---
# prd
## 什么时候使用
- 把工程方向的常用动作沉淀成 Agent 可调用的技能 适合处理工程开发场景下的代码实现、调试、重构、测试或代码审查，核心价值是把输入、判断、执行、验证和交付边界固定下来，避免 Agent 泛泛回答。 把任务拆成可执行、可检查、可继续迭代…
- 面向代码实现、重构、调试或代码审查，优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Step 1: Gather Context / Step 2: Write the PRD / Problem Statement」组织步骤，不把推断写成作者事实。
- 读取文件、写入/修改文件；主要在本地完成；通常不需要额外 API Key。
- 先跑小样例，确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown，不编造命令、平台或依赖。 证据边界与执行链路
作者原文负责流程事实；仓库文件负责来源和命令；流狐只补充适用场景、限制和质量判断。
skill "prd" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Step 1: Gather Context / Step 2: Write the PRD / Problem Statement
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
}Product Requirements Document
Write a PRD that answers two questions clearly: what are we building and why does it matter. Everything else is supporting detail.
Step 1: Gather Context
Before writing anything, nail down these three inputs. If the user hasn't provided them, ask.
| Input | Question | Watch out for |
|---|---|---|
| Problem | What pain exists today? Who feels it? | Jumping to solutions before the problem is sharp |
| Audience | Who specifically benefits? What's their situation? | "Everyone" is not an audience |
| Current state | How do people solve this now? What's broken? | Assuming nothing exists today |
If a JTBD analysis, research, or prior conversations exist, pull from them. Don't invent context.
Step 2: Write the PRD
Use this structure. Every section earns its place — skip a section only if it genuinely doesn't apply.
# PRD: [Feature / Product Name]
## Problem Statement
What's broken, missing, or painful. Ground it in real user behavior.
Who has this problem. How often. How bad.
## Goals & Non-Goals
### Goals
- [ ] Measurable outcome this work achieves
- [ ] Another measurable outcome
### Non-Goals
- What this work deliberately does NOT try to solve (and why)
## User Stories / Job Stories
Use job story format when possible:
- When [situation], I want to [motivation], so I can [outcome]
Keep to 3-8 stories. If you have more, the scope is too big.
## Functional Requirements
What the system must do. Number them for traceability.
FR-1: [Requirement]
FR-2: [Requirement]
## Non-Functional Requirements
Performance, security, accessibility, scale, compatibility.
NFR-1: [Requirement]
NFR-2: [Requirement]
## Success Metrics
How we'll know this worked. Be specific:
- Metric + target + timeframe
- e.g., "Reduce average onboarding time from 12 min to under 5 min within 30 days of launch"
## Open Questions
What's unresolved. Who owns answering it. When it needs an answer by.
| # | Question | Owner | Deadline |
|---|----------|-------|----------|
## Out of Scope
What's explicitly excluded from this effort. Prevents scope creep later.Quality Checks
Run these before delivering the PRD:
- Every requirement is testable. If you can't write a pass/fail check for it, rewrite it. "Fast" is not testable. "Loads in under 2 seconds on 3G" is.
- No solutions hiding as requirements. "Use Redis for caching" is a solution. "Cache frequently accessed data with sub-100ms retrieval" is a requirement.
- Scope is realistic. If the PRD has 30+ requirements, it's a roadmap pretending to be a PRD. Split it.
- Goals are measurable. Each goal should have a number attached or a clear yes/no test.
- Non-goals are deliberate. They prevent future arguments about what "should have been included."
- Open questions have owners. An open question without an owner is a stalled decision.
Tone and Format
- Write requirements as imperative statements: "The system shall..." or "Users can..."
- Use numbered identifiers (FR-1, NFR-1) so engineers and designers can reference them
- Keep the language concrete. Replace "intuitive" with observable behavior. Replace "scalable" with specific numbers.
- One requirement per line. Compound requirements hide complexity.
When the PRD Is Done
Hand it off with three things clear:
- What's decided — requirements that are locked
- What's open — questions that still need answers
- What's next — who reviews, who builds, what's the timeline
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核