Token导航 LogoToken导航

Agent Skill · 研究检索

requirement-workflow(需求工作流程)

requirement-workflow 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

requirement-workflow 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,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 39,733。

总安装 1,371周安装 56GitHub Stars 1下载 444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:requirement-workflow(需求工作流程)
来源仓库:https://github.com/learnwy/skills
仓库路径:skills/requirement-workflow
安装命令:
npx skills add https://github.com/learnwy/skills --skill requirement-workflow
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/learnwy/skills --skill requirement-workflow

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。

requirement-workflow 用于查找、检索和筛选相关信息。它适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词、任务场景或来源线索快速定位候选结果。可结合来源仓库、安装命令和原始 README 继续核验具体用法。

安装命令为 npx skills add https://github.com/learnwy/skills --skill requirement-workflow。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/requirement-workflow

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/learnwy/skills --skill requirement-workflow;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

requirement-workflow 怎么安装?

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

requirement-workflow 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。

requirement-workflow 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。

requirement-workflow 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。

requirement-workflow 用的人多吗?

公开数据是累计安装 1,371 次,近一周安装 56 次,下载 444 次,GitHub 1 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。

requirement-workflow 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

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。

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。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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
requirement-workflow

需求工作流程 · 本页

1,3711需要联网未明确有提醒

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

Requirement Workflow (SDD)

Structured development orchestrator based on Spec-Driven Development: define specs first, decompose into tasks, execute with quality gates, verify against acceptance criteria.

Core principle: Spec is the single source of truth. Code follows spec, not the other way around.

Prerequisites

  • Node.js >= 18
  • Writable project directory for workflow state files (.trae/workflow/)

Quick Start

User says anything about building/implementing/fixing →
  1. AI classifies the request (type + size + risk)
  2. AI runs: node init.cjs -r <root> -n <name> -t <type> -s <size> -k <risk>
  3. AI fills spec.md with structured requirements (EARS format)
  4. AI decomposes spec.md into tasks.md
  5. AI implements task by task, advancing stages
  6. AI verifies against checklist.md

Working Modes

This skill supports two modes, selected based on classification:

Agent Mode (Default)

For small/tiny scope or bugfix type: skip spec phase, go straight to implementation.

INIT → IMPLEMENTING → TESTING → DONE

Spec Mode (SDD)

For medium/large scope or elevated/critical risk: full SDD lifecycle.

INIT → DEFINING → PLANNING → DESIGNING → IMPLEMENTING → TESTING → DELIVERING → DONE

Classification Matrix

SignalTypeSizeRisk
"fix bug", "broken", "crash"bugfixtiny-smallnormal
"add feature", "implement", "build"featuresmall-largenormal
"refactor", "clean up", "reorganize"refactorsmall-mediumnormal
"tech debt", "upgrade", "migrate"tech-debtmedium-largeelevated
auth, payments, data, securityanyanyelevated-critical
>15 files estimatedanylargeelevated

Size Heuristics

SizeFilesDurationStages
tiny≤2<30minINIT → IMPLEMENTING → DONE
small3-530min-2hINIT → IMPLEMENTING → TESTING → DONE
medium6-152h-1dFull SDD
large>15>1dFull SDD + all checkpoints

SDD Lifecycle

Stage 1: INIT

Classify request, initialize workflow, create artifact directory.

node init.cjs -r <project_root> -n "<name>" -t <type> -s <size> -k <risk>

Output: workflow.yaml, empty spec.md, tasks.md, checklist.md

Stage 2: DEFINING (Spec Mode only)

Fill spec.md with structured requirements using EARS format:

# Feature Name

## Background
{Why this is needed — problem statement}

## Scope
- In: {what IS included}
- Out: {what is NOT included}

## Acceptance Criteria (EARS format)
- [ ] When <condition>, the system shall <response>
- [ ] While <state>, the system shall <behavior>
- [ ] Where <constraint>, the system shall <limit>

## Constraints
- {Performance, security, compatibility requirements}

## Out of Scope
- {Explicitly deferred items}

Checkpoint: If risk is elevated/critical, pause for user review of spec.md.

Hooks: pre_stage_DEFINING → iron-audit-pm, problem-definer, risk-auditor

Stage 3: PLANNING (Spec Mode only)

Decompose spec.md into tasks.md — every acceptance criterion maps to ≥1 task:

# Tasks

## Phase 1: Foundation
- [ ] Task 1.1: {description} [files: x, y]
- [ ] Task 1.2: {description} [files: z]

## Phase 2: Core Logic
- [ ] Task 2.1: {description} [files: a, b]

## Phase 3: Integration & Polish
- [ ] Task 3.1: {description} [files: c]

## Verification
- [ ] All acceptance criteria pass
- [ ] Lint clean
- [ ] Type check pass

Rule: Each task must be atomic (completable independently) and traceable to a spec item.

Hooks: pre_stage_PLANNING → story-mapper, mvp-freeze-architect

Stage 4: DESIGNING (Spec Mode only)

Create design.md (only for medium+ size) with architecture decisions:

  • Component structure
  • Data flow
  • API contracts
  • Key trade-offs

Checkpoint: If risk is elevated/critical, pause for user review.

Hooks: pre_stage_DESIGNING → domain-modeler, architecture-advisor, responsibility-modeler

Stage 5: IMPLEMENTING

Execute tasks from tasks.md sequentially. For each task:

  1. Read the task description
  2. Implement the change
  3. Mark task as [x] in tasks.md
  4. Run relevant tests if available

Hooks: pre_stage_IMPLEMENTING → tdd-coach

Stage 6: TESTING

Run full test suite. Update checklist.md:

# Checklist

## Code Quality
- [ ] Implementation complete
- [ ] Lint clean (run lint command)
- [ ] Type check pass (run typecheck command)

## Tests
- [ ] Unit tests pass
- [ ] Integration tests pass (if applicable)

## Acceptance Criteria
- [ ] AC 1: {criterion from spec} — verified
- [ ] AC 2: {criterion from spec} — verified

## Review
- [ ] Self-review complete
- [ ] No TODO/FIXME left unresolved

Hooks: pre_stage_TESTING → test-strategy-advisor, test-strategist, code-reviewer

Stage 7: DELIVERING (Spec Mode only)

Final verification against spec.md. Ensure every acceptance criterion has been met.

Hooks: post_stage_DELIVERING → code-reviewer, tech-design-reviewer

Stage 8: DONE

Workflow complete. Summary output:

  • What was delivered
  • Files changed
  • Tests passed
  • Time elapsed

Stage Advancement

# Check current status
node status.cjs -r <project_root>

# Advance to next stage
node advance.cjs -r <project_root>

# Force advance (skip checkpoint)
node advance.cjs -r <project_root> --force

Hooks System

Hooks are agents/skills that run at stage transitions. Three scopes:

ScopeFilePriority
Globalhooks.yaml (skill dir)Lowest
Project.trae/workflow/hooks.yamlMedium
Workflowworkflow.yaml (in workflow dir)Highest

Hook Points

HookWhenDefault Agents
pre_stage_DEFININGBefore filling speciron-audit-pm, risk-auditor
pre_stage_PLANNINGBefore task decompositionstory-mapper, mvp-freeze-architect
pre_stage_DESIGNINGBefore architecturedomain-modeler, architecture-advisor
pre_stage_IMPLEMENTINGBefore codingtdd-coach
pre_stage_TESTINGBefore test phasetest-strategy-advisor, code-reviewer
post_stage_DELIVERINGAfter final checktech-design-reviewer
# List hooks
node hooks.cjs -r <project_root> list

# Add a hook
node hooks.cjs -r <project_root> add pre_stage_TESTING -n my-validator --type skill

Quality Gates

Checkpoint Rules

StageCheckpoint When
DEFININGrisk = elevated or critical
PLANNINGsize = large or risk ≥ elevated
DESIGNINGsize ≥ medium or risk ≥ elevated
TESTINGalways (all risk levels)

At checkpoints, AI pauses and asks user for confirmation before advancing.

SDD Traceability Rule

Every line of code must trace back to:

  1. A task in tasks.md
  2. Which traces to an acceptance criterion in spec.md

If you find yourself writing code not covered by any task — stop and update tasks.md first.

Error Handling

IssueSolution
User gives vague requestClassify as feature/small, use Agent Mode, refine during implementation
Spec is incompleteAdd missing acceptance criteria before advancing to PLANNING
Task is too largeBreak into sub-tasks, each ≤1 file change
Tests fail during TESTINGStay in TESTING, fix issues, re-run
Checkpoint rejected by userStay in current stage, revise artifacts per feedback
Workflow abandonedNo cleanup needed, state persists in .trae/workflow/

Scripts

ScriptPurpose
init.cjsInitialize workflow with classification
advance.cjsAdvance to next stage
status.cjsShow current workflow status
hooks.cjsManage hook registrations
lib/common.cjsShared YAML utilities

Agents

See AGENTS.md for the full registry. Key agents by phase:

PhaseAgentMethodology
DEFININGiron-audit-pmPRD audit, DNA extraction
DEFININGrisk-auditorRisk scanning
DEFININGproblem-definerWeinberg problem analysis
PLANNINGstory-mapperPatton story mapping
PLANNINGmvp-freeze-architectScope freezing
DESIGNINGdomain-modelerDDD/Evans modeling
DESIGNINGarchitecture-advisorQuality attributes
IMPLEMENTINGtdd-coachBeck TDD cycle
TESTINGtest-strategistCrispin test strategy
TESTINGcode-reviewerCode review
DELIVERINGtech-design-reviewerArchitecture review

References

平台分布

Codex

34.6%
按下载量换算154

Claude

29.37%
按下载量换算130

Cursor

19.28%
按下载量换算86

Gemini CLI

9.69%
按下载量换算43

安全审计

Gen Agent Trust Hub

通过

Socket

可疑

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。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