Token导航 LogoToken导航

Agent Skill · 研究检索

spec-writing(规范编写)

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。

spec-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,权限为接触敏感数据;许可证为 MIT,暂无第三方扫描结果,按安装量在研究检索类的 73,675 个 Skill 里排第 69,087。

总安装 220周安装 9GitHub Stars 公开资料未说明下载 71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
MIT

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:spec-writing(规范编写)
来源仓库:https://github.com/linaqruf/cc-plugins
仓库路径:skills/spec-writing
安装命令:
npx skills add linaqruf/cc-plugins --skill "spec-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。

AgentSkills.tonpx skills
npx skills add linaqruf/cc-plugins --skill "spec-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。

spec-writing 用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写。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。通过 npx skills add linaqruf/cc-plugins --skill "spec-writing" 命令安装。

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。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/spec-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 linaqruf/cc-plugins --skill "spec-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

    它需要接触敏感数据,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

spec-writing 怎么安装?

在终端执行 npx skills add linaqruf/cc-plugins --skill "spec-writing"。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

spec-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。

spec-writing 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

spec-writing 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

spec-writing 用的人多吗?

公开数据是累计安装 220 次,近一周安装 9 次,下载 71 次。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。

spec-writing 安全吗?

没有查到第三方扫描结果,安全性无法从公开资料判断,安装前请自行审查源码、依赖和命令行为。

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。
  • 许可证明确为 MIT,商用和分发边界好判断。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 没有第三方扫描结果,安全性只能靠自己读源码判断。
  • 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

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 要求有安全扫描结论才肯安装的团队。
  • 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
spec-writing

规范编写 · 本页

220—接触敏感数据MIT未收录

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

name
spec-writing
description
Use when the user wants to create project specs, design systems, or feature plans. Triggers on "create spec", "plan project", "design system", "plan feature", or "write specification".
version
3.0.0

Spec Writing v3.0

Generate comprehensive project specifications with SPEC.md as the core file and optional supplements for reference material.

Core Principle

SPEC.md is always the complete spec. SPEC/ files are optional lookup supplements.

SPEC.md               # Always created, always self-sufficient
CLAUDE.md             # Generated with spec references

SPEC/                 # Optional, created when user agrees
├── api-reference.md  # Lookup: endpoint schemas, request/response
├── sdk-patterns.md   # Lookup: external SDK usage patterns
└── data-models.md    # Lookup: complex entity schemas

Key distinction:

  • SPEC.md = Things you READ (narrative, decisions, requirements)
  • **SPEC/*.md** = Things you LOOK UP (schemas, SDK patterns, external API details)

Single Adaptive Flow

One interview flow replaces previous Quick/SPEC/DEEP modes:

Interview starts
    ↓
Build SPEC.md progressively
    ↓
Hit reference-heavy topic? ──→ Ask: "Create SPEC/[topic].md for lookup?"
    ↓                                    ↓
Continue interview                  User decides (yes/no)
    ↓
Generate SPEC.md + CLAUDE.md
    ↓
If user said yes → Generate SPEC/[topic].md files

Interview Workflow

Phase 1: Vision & Problem

  • Problem statement
  • Target users
  • Success criteria

Phase 2: Requirements

  • MVP features (must-have)
  • Out of scope (explicit)
  • User flows

Phase 3: Architecture

  • System type detection
  • Architecture pattern (present 2-3 alternatives with tradeoffs)
  • Tech stack with recommendations

Phase 4: Tech Stack Details

  • Frontend (if applicable)
  • Backend (if applicable)
  • Database (if applicable)
  • Use multiple choice with recommendations

Phase 5: Design & Security

  • Visual design (if frontend)
  • Authentication approach
  • Constraints & compliance

Supplement Prompts (Mid-Interview)

When hitting reference-heavy topics, ask:

"Your API has 15 endpoints with detailed schemas. Should I: - A) Keep it inline in SPEC.md (shorter reference section) - B) Create SPEC/api-reference.md as a separate lookup file"

Create supplements only for:

  • Reference material - Stuff you look up, not read through
  • External dependencies - SDK docs, library patterns, third-party APIs

Opinionated Recommendations

Lead with recommended options, allow override:

Which package manager?

- A) bun (Recommended) - Fastest, built-in test runner, drop-in npm replacement
- B) pnpm - Fast, strict dependency resolution, good for monorepos
- C) npm - Universal compatibility, no setup needed
- D) yarn - If team already uses it

Principles:

  1. Lead with recommended option + brief rationale
  2. Context-aware (desktop app? acknowledge Tauri vs Electron tradeoffs)
  3. Acknowledge "it depends" cases (team familiarity, existing codebase)
  4. Stay current with ecosystem changes

SPEC.md Structure

# [Project Name]

## Overview
Problem, solution, target users, success criteria.

## Product Requirements
Core features (MVP), future scope, out of scope, user flows.

## Technical Architecture
Tech stack (with rationale), system design diagram, data models, API endpoints.

## System Maps
- Architecture diagram (ASCII)
- Data model relations
- User flow diagrams
- Wireframes (key screens)

## Design System
(If frontend) Colors, typography, components, accessibility.

## File Structure
Project directory layout.

## Development Phases
Phased implementation plan with checkboxes.

## Open Questions
Decisions to make during development.

---

## References
(If supplements exist) Trigger-based links to SPEC/ files.

Connecting SPEC.md to Supplements

When supplements exist, reference them with triggers:

Inline (in relevant sections):

## API Design

**Endpoints overview:**
- `POST /auth/login` - User authentication
- `GET /projects` - List user projects

→ When implementing endpoints, reference `SPEC/api-reference.md` for full request/response schemas.

References section (bottom):

---

## References

→ When implementing API endpoints: `SPEC/api-reference.md`
→ When using Anthropic SDK: `SPEC/sdk-patterns.md`

CLAUDE.md Generation

Agent-optimized pointer file:

# [Project Name]

[One-line description]

## Spec Reference

Primary spec: `SPEC.md`

→ When implementing API endpoints: `SPEC/api-reference.md`
→ When using [SDK/Library]: `SPEC/sdk-patterns.md`

## Key Constraints

- [Critical constraint 1 - surfaced from spec]
- [Critical constraint 2]
- [Out of scope reminder]

## Commands

- `[package-manager] run dev` - Start development
- `[package-manager] run test` - Run tests
- `[package-manager] run build` - Production build

## Current Status

→ Check `SPEC.md` → Development Phases section

Principles:

  • Surface critical constraints directly (prevent missed context)
  • Trigger-based supplement references
  • Short - pointer, not duplication
  • Status points to SPEC.md (single source)

Context7 Integration

After tech choices, fetch relevant documentation:

1. resolve-library-id for each technology
2. query-docs for setup guides and patterns
3. Include insights in relevant spec files

Best Practices

Interview Conduct

  • Multiple choice: Use AskUserQuestion options, not open-ended text
  • 2-3 alternatives: For key decisions, show options with tradeoffs
  • YAGNI: Ruthlessly simplify - "Do we really need this for MVP?"
  • Supplements on demand: Only offer when content is truly reference-heavy

Output Quality

  • Be specific and actionable
  • Include code examples for data models
  • Reference Context7 documentation
  • Keep scope realistic for MVP
  • Include system maps (architecture, data relations, user flows)

Reference Files

Templates

  • references/output-template.md - SPEC.md structure
  • templates/index.template.md
  • templates/overview.template.md
  • templates/architecture.template.md
  • templates/frontend.template.md
  • templates/backend.template.md
  • templates/design-system.template.md
  • templates/api-reference.template.md
  • templates/cli-reference.template.md
  • templates/data-models.template.md
  • templates/security.template.md

References

  • references/interview-questions.md - Question bank with recommendations
  • references/spec-folder-template.md - Supplement structure guide

Examples

  • examples/web-app-spec.md
  • examples/cli-spec.md
  • examples/api-spec.md
  • examples/library-spec.md

Related Commands

  • /spec - Generate project specification
  • /feature - Generate feature specification
  • /design - Generate design system specification
  • /sync - Sync spec with codebase changes (git-aware)

Integration with Other Skills

feature-dev

After creating specs, use feature-dev agents:

  1. code-explorer - Analyze existing patterns
  2. code-architect - Design implementation
  3. code-reviewer - Review implementation

frontend-design

Use design specs to implement components following the specification.

平台分布

Codex

27.48%
按下载量换算20

windsurf

23.84%
按下载量换算17

OpenCode

20.69%
按下载量换算15

Claude Code

12.28%
按下载量换算9

Antigravity

8.92%
按下载量换算6

Gemini CLI

3.39%
按下载量换算2

安全审计

暂无安全审计结果可展示。

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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