Token导航 LogoToken导航

Agent Skill · 研究检索

n-version-workflow(n 版本工作流程)

n-version-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。

n-version-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,权限为执行命令;许可证为 MIT,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 32,827。

总安装 2,448周安装 102GitHub Stars 55下载 816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:n-version-workflow(n 版本工作流程)
来源仓库:https://github.com/rysweet/amplihack
仓库路径:skills/n-version-workflow
安装命令:
npx skills add https://github.com/rysweet/amplihack --skill n-version-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。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/rysweet/amplihack --skill n-version-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。

n-version-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。适用于版本控制、工作流程管理和多版本维护场景。

通过关键词搜索、来源仓库和原始 README 核验具体用法,结合安装命令使用。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。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/n-version-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/rysweet/amplihack --skill n-version-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

    它需要执行命令,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

n-version-workflow 怎么安装?

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

n-version-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。

n-version-workflow 需要什么权限?

权限级别是执行命令。安装流程涉及命令执行,可能通过 npx skills add https://github.com/rysweet/amplihack --skill n-version-workflow;npx skills add rysweet/amplihack --skill "n-version-workflow" 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

n-version-workflow 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。

n-version-workflow 用的人多吗?

公开数据是累计安装 2,448 次,近一周安装 102 次,下载 816 次,GitHub 55 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。

n-version-workflow 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。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 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

  • 需要执行命令,装到常用环境前要先评估影响面。
  • 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。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 不能接受 Skill 在你机器上执行命令的人。
  • 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
n-version-workflow

n 版本工作流程 · 本页

2,44855执行命令MIT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

N-Version Programming Workflow Skill

Purpose

Execute N-version programming workflow for critical implementations where multiple independent solutions should be generated and compared to select the best approach.

When to Use This Skill

USE FOR:

  • Critical security features (authentication, authorization)
  • Complex algorithms with multiple valid approaches
  • High-risk refactoring of core components
  • Architecture decisions with significant long-term impact
  • When correctness is paramount over speed

AVOID FOR:

  • Simple CRUD operations
  • Straightforward bug fixes
  • Documentation updates
  • Minor UI tweaks
  • Time-sensitive quick fixes

Configuration

Core Parameters

N (Number of Versions):

  • 3 - Default for standard tasks
  • 4-6 - Critical features requiring high confidence
  • 2 - Quick validation of approach

Selection Criteria (priority order):

  1. Correctness - Meets requirements and passes tests
  2. Security - No vulnerabilities or anti-patterns
  3. Simplicity - Ruthless simplicity, minimal complexity
  4. Philosophy Compliance - Follows project principles
  5. Performance - Efficiency and resource usage

Agent Diversity Profiles:

  • conservative - Proven patterns and safety
  • innovative - Novel approaches and optimizations
  • minimalist - Ruthless simplicity
  • pragmatic - Balance trade-offs for practical solutions
  • performance-focused - Optimize for speed and efficiency

Execution Process

Step 1: Prepare Common Context

  • Use prompt-writer agent to create crystal-clear specification
  • Document all requirements explicitly
  • Define success criteria measurably
  • Prepare identical task specification for all N versions
  • Identify evaluation metrics upfront
  • CRITICAL: Capture explicit user requirements that CANNOT be optimized away

Output: Single authoritative specification document

Step 2: Generate N Independent Implementations

  • Spawn N Claude subprocesses simultaneously
  • Each subprocess receives IDENTICAL task specification
  • NO context sharing between subprocesses (true independence)
  • Each uses different agent diversity profile
  • Each produces complete implementation with tests
  • Each works in isolated directory (version_1/, version_2/, etc.)

Example for N=3:

  • Subprocess 1: Conservative approach (proven patterns, comprehensive error handling)
  • Subprocess 2: Pragmatic approach (balance simplicity and robustness)
  • Subprocess 3: Minimalist approach (ruthless simplification, minimal dependencies)

Step 3: Collect and Compare Implementations

  • Wait for all N implementations to complete
  • Use analyzer agent to examine each implementation
  • Use tester agent to run tests for each version
  • Document results in comparison matrix

Comparison Matrix:

| Version | Correctness | Security | Simplicity | Philosophy | Performance | Lines of Code |
|---------|-------------|----------|------------|------------|-------------|---------------|
| v1      | PASS        | PASS     | 7/10       | 8/10       | 150ms       | 180           |
| v2      | PASS        | PASS     | 9/10       | 9/10       | 180ms       | 95            |
| v3      | FAIL        | N/A      | 10/10      | 7/10       | N/A         | 65            |

Step 4: Review and Evaluate

  • Use reviewer agent for comprehensive comparison
  • Use security agent to evaluate security of each version
  • Apply selection criteria in priority order
  • Eliminate versions that fail correctness tests
  • Compare remaining versions on other criteria
  • Identify best parts of each implementation

Evaluation Process:

  1. Filter: Remove versions failing correctness tests
  2. Security Gate: Eliminate versions with security issues
  3. Philosophy Check: Score each on simplicity and compliance
  4. Performance Compare: Measure and compare benchmarks
  5. Synthesis: Identify if hybrid approach could be superior

Step 5: Select or Synthesize Solution

Decision Tree:

  1. Is there ONE version that passes all criteria?

- YES → Select it and document why - NO → Continue to step 2

  1. Are there 2+ versions tied on top criteria?

- YES → Continue to step 3 - NO → Select highest scoring version

  1. Do versions have complementary strengths?

- YES → Synthesize hybrid combining best parts - NO → Select based on weighted criteria priority

Example Synthesis:

Selected: Hybrid of v1 and v2
- Core logic from v2 (ruthless simplicity)
- Error handling from v1 (comprehensive coverage)
- Testing approach from v2 (focused, minimal)
- Documentation style from v1 (thorough)

Rationale: v2's minimalist core paired with v1's robust
error handling provides optimal balance of simplicity
and production-readiness.

Step 6: Implement Selected Solution

  • Use builder agent to implement final version
  • If single version selected: Use it directly
  • If synthesis: Implement hybrid combining best parts
  • Preserve all explicit user requirements from Step 1
  • Run full test suite
  • Document selection rationale in code comments

Documentation Template:

"""
N-Version Implementation Selection

Generated Versions: 3
Selection: Hybrid of v1 (conservative) and v2 (pragmatic)

Rationale:
- v1 had superior error handling and edge case coverage
- v2 had cleaner architecture and better testability
- v3 failed correctness tests (edge case handling)

This implementation combines v2's core logic with v1's
defensive programming approach for production robustness.

Selection Criteria Applied:
1. Correctness: v1=PASS, v2=PASS, v3=FAIL
2. Security: All passed
3. Simplicity: v2 ranked highest
4. Philosophy: v1 and v2 tied
5. Performance: Negligible difference

Step 7: Document Learnings

  • Create analysis document: n_version_analysis.md
  • Document all N implementations generated
  • Explain selection rationale in detail
  • Capture insights from rejected versions
  • Store patterns learned in memory using store_discovery() from amplihack.memory.discoveries
  • Include comparison matrix for future reference

Trade-Offs

Cost: N times the compute resources and time Benefit: Significantly reduced risk of critical errors Best For: Features where bugs are expensive (security, data integrity)

Examples

Example 1: Authentication System

Task: Implement JWT-based authentication Configuration: N=4 (critical security feature) Profiles: conservative, security-focused, pragmatic, minimalist

Result:

  • v1 (conservative): Most comprehensive but over-engineered
  • v2 (security-focused): Excellent security but complex
  • v3 (pragmatic): Good balance, missing edge cases
  • v4 (minimalist): Too simple, security gaps

Selection: Hybrid of v2 and v3

  • Security implementation from v2
  • API design and simplicity from v3

Rationale: Security cannot be compromised, but v3's cleaner API design improved usability without sacrificing security.

Example 2: Data Processing Pipeline

Task: Process large CSV files efficiently Configuration: N=3 (performance-critical) Profiles: pragmatic, performance-focused, minimalist

Result:

  • v1 (pragmatic): Pandas-based, familiar but slow
  • v2 (performance-focused): Custom streaming, 10x faster
  • v3 (minimalist): Python CSV module, simple but slow

Selection: v2 (performance-focused)

Rationale: Performance requirements justified complexity. v2's streaming approach met throughput requirements while v1 and v3 could not scale.

Philosophy Alignment

This workflow enforces:

  • Reduced Risk: Multiple implementations catch errors single approach might miss
  • Exploration: Different approaches reveal design trade-offs
  • Evidence-Based Selection: Systematic comparison vs. gut feeling
  • Learning: Rejected versions still provide valuable insights
  • Parallel Execution: N implementations run simultaneously for efficiency

Integration with Default Workflow

This workflow replaces Steps 4-5 (Research/Design and Implementation) of the DEFAULT_WORKFLOW when enabled. All other steps (requirements, testing, CI/CD) remain the same.

平台分布

Claude Code

28.27%
按下载量换算231

Antigravity

24.25%
按下载量换算198

OpenCode

15.99%
按下载量换算130

Gemini CLI

11.72%
按下载量换算96

windsurf

7.53%
按下载量换算61

Cursor

3.35%
按下载量换算27

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/rysweet/amplihack --skill n-version-workflow;npx skills add rysweet/amplihack --skill "n-version-workflow" 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

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。

继续浏览同类 Skills