Token导航 LogoToken导航

Agent Skill · 开发

ralph-loop-writer(拉尔夫·卢普 编剧)

ralph-loop-writer 用于补充开发相关能力,适合在 OpenClaw 中需要让 Agent 承接开发相关任务时使用。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。

ralph-loop-writer 是开发类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为执行命令;许可证为 MIT-0,第三方扫描有提醒项,按安装量在开发类的 21,746 个 Skill 里排第 894。

总安装 51,888周安装 2,162GitHub Stars 3下载 17,296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-0
版本
1.0.0

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:ralph-loop-writer(拉尔夫·卢普 编剧)
来源仓库:https://github.com/walkamolee/ralph-loop-writer
安装命令:
openclaw skills install ralph-loop-writer
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 OpenClaw 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。

ClawHubOpenClaw
openclaw skills install ralph-loop-writer

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

为 Claude Code、Gemini CLI 或 Grok CLI 生成可立即运行的 shell 循环命令,并通过 AI 工具、模型、shell、复杂性和循环功能进行自定义。

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 OpenClaw;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 openclaw skills install ralph-loop-writer;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

ralph-loop-writer 怎么安装?

在终端执行 openclaw skills install ralph-loop-writer。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

ralph-loop-writer 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。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。

ralph-loop-writer 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 openclaw skills install ralph-loop-writer 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。

ralph-loop-writer 是免费的吗?

来源标注的许可证是 MIT-0。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。

ralph-loop-writer 用的人多吗?

公开数据是累计安装 51,888 次,近一周安装 2,162 次,下载 17,296 次,GitHub 3 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。

ralph-loop-writer 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

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。

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 OpenClaw,想直接扩展它能力的人。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

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。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
ralph-loop-writer

拉尔夫·卢普 编剧 · 本页

51,8883执行命令MIT-0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

name
ralph
description
Generate Claude Code, Gemini CLI, or Grok CLI automation loop commands. Asks questions about your requirements and outputs a ready-to-run command for PowerShell, Windows CMD, or Bash/Linux.
allowed-tools

Ralph Command Generator

Generate optimized loop commands for automating Claude Code, Gemini CLI, or Grok CLI with PROMPT.md.

Step 1: Choose AI Tool

Use AskUserQuestion:

  • Question: "Which AI CLI tool do you want to use?"
  • Header: "AI Tool"
  • Options:

1. "Claude Code (Recommended)" - "Claude AI assistant CLI" 2. "Gemini CLI" - "Google Gemini AI assistant" 3. "Grok CLI" - "xAI Grok AI assistant with agentic coding"

Store the choice for later.

Step 2: Choose Model

Use AskUserQuestion based on AI tool choice:

If Claude Code selected:

  • Question: "Which Claude model do you want to use?"
  • Header: "Model"
  • Options:

1. "Default (Recommended)" - "Use default model (currently Sonnet 4.5)" 2. "Haiku" - "Fastest and most cost-effective" 3. "Sonnet" - "Balanced performance and cost" 4. "Opus" - "Most capable, higher cost"

If Gemini CLI selected:

  • Question: "Which Gemini model do you want to use?"
  • Header: "Model"
  • Options:

1. "Default (Recommended)" - "Use default model" 2. "gemini-3-flash" - "Latest Gemini 3, fastest and most cost-effective" 3. "gemini-3-pro" - "Latest Gemini 3, most capable for complex tasks" 4. "gemini-2.5-flash" - "Stable production model, fast" 5. "gemini-2.5-pro" - "Stable production model, more capable"

If Grok CLI selected:

  • Question: "Which Grok model do you want to use?"
  • Header: "Model"
  • Options:

1. "Default (Recommended)" - "Use grok-code-fast-1, optimized for fast code generation and agentic loops" 2. "grok-4-latest" - "Latest Grok 4, most capable for complex reasoning" 3. "grok-beta" - "Preview of upcoming features"

Store the choice for later.

Step 3: Choose Operating System

Use AskUserQuestion:

  • Question: "Which shell environment are you using?"
  • Header: "Shell"
  • Options:

1. "PowerShell (Recommended)" - "Windows PowerShell" 2. "Windows CMD" - "Command Prompt (cmd.exe)" 3. "Bash/Linux" - "Linux, Mac, or WSL"

Store the choice for later.

Step 4: Choose Complexity Level

Use AskUserQuestion:

  • Question: "What level of control do you need?"
  • Header: "Complexity"
  • Options:

1. "Simple (Recommended)" - "Basic loop with minimal options" 2. "Intermediate" - "Combine 2 control mechanisms" 3. "Advanced" - "Full control with multiple safeguards"

Step 5: Choose Loop Type (Based on Complexity)

If Simple:

Use AskUserQuestion:

  • Question: "What type of loop?"
  • Header: "Loop Type"
  • Options:

1. "Fixed count (Recommended)" - "Run exactly N times" 2. "Infinite with delay" - "Run forever with pauses" 3. "Stop file trigger" - "Run until STOP.txt exists"

If Intermediate:

Use AskUserQuestion:

  • Question: "What combination do you need?"
  • Header: "Features"
  • multiSelect: true
  • Options:

1. "Fixed iterations" - "Run max N times" 2. "Time limit" - "Run for max X minutes" 3. "Stop file" - "Stop when STOP.txt appears" 4. "Delay between runs" - "Pause X seconds" 5. "Show counter" - "Display run number" 6. "File monitoring" - "Stop at file size/lines"

If Advanced:

Use AskUserQuestion:

  • Question: "What features do you want?"
  • Header: "Features"
  • multiSelect: true
  • Options:

1. "Max iterations (Recommended)" - "Limit runs" 2. "Time limit (Recommended)" - "Max duration" 3. "Stop file (Recommended)" - "Manual stop" 4. "Delay between runs" - "Pause X seconds" 5. "Timestamp logging" - "Show time of each run" 6. "Counter display" - "Show run number"

Step 6: Gather Parameters

Based on selected features, ask for values:

If fixed iterations selected:

  • Question: "How many iterations?"
  • Options: "5 (Recommended)", "10", "20", "50", "100", "Custom"

If time limit selected:

  • Question: "Maximum duration?"
  • Options: "10 minutes (Recommended)", "30 minutes", "1 hour", "Custom"

If delay selected:

  • Question: "Delay between runs?"
  • Options: "5 seconds (Recommended)", "10 seconds", "30 seconds", "Custom"

If file monitoring selected:

  • Question: "Monitor by?"
  • Options: "File size (e.g., 5KB)", "Line count (e.g., 50 lines)", "Content (e.g., 'THE END')"

Step 7: Generate Command

Build the appropriate command based on:

  1. AI tool choice (Claude or Gemini)
  2. Model choice
  3. Shell choice (PowerShell, CMD, Bash)
  4. Complexity level
  5. Selected features
  6. Parameter values

IMPORTANT - Command Syntax:

For Claude Code (PowerShell/Bash): Use claude-code (NOT claude -p) to accept piped input. The -p flag requires an argument, not pipe.

  • Default: Get-Content PROMPT.md -Raw | claude-code --dangerously-skip-permissions
  • Haiku: Get-Content PROMPT.md -Raw | claude-code --model haiku --dangerously-skip-permissions
  • Sonnet: Get-Content PROMPT.md -Raw | claude-code --model sonnet --dangerously-skip-permissions
  • Opus: Get-Content PROMPT.md -Raw | claude-code --model opus --dangerously-skip-permissions

For Claude Code (Bash):

  • Default: cat PROMPT.md | claude-code --dangerously-skip-permissions
  • With model: cat PROMPT.md | claude-code --model haiku --dangerously-skip-permissions

For Gemini CLI (PowerShell): Gemini CLI accepts stdin piping.

  • Default: Get-Content PROMPT.md -Raw | gemini --yolo
  • gemini-3-flash: Get-Content PROMPT.md -Raw | gemini --model gemini-3-flash --yolo
  • gemini-3-pro: Get-Content PROMPT.md -Raw | gemini --model gemini-3-pro --yolo
  • gemini-2.5-flash: Get-Content PROMPT.md -Raw | gemini --model gemini-2.5-flash --yolo
  • gemini-2.5-pro: Get-Content PROMPT.md -Raw | gemini --model gemini-2.5-pro --yolo

For Grok CLI (PowerShell):

  • Default: Get-Content PROMPT.md -Raw | grok-auto (uses default model from GROK_MODEL env var, auto-approves permissions)
  • grok-code-fast-1: Get-Content PROMPT.md -Raw | grok-auto -m grok-code-fast-1
  • grok-4-latest: Get-Content PROMPT.md -Raw | grok-auto -m grok-4-latest
  • grok-beta: Get-Content PROMPT.md -Raw | grok-auto -m grok-beta

Command Templates

IMPORTANT - Placeholder Replacement:

For Claude Code: Replace [AI_COMMAND_WITH_PROMPT] with the FULL command including the prompt argument:

  • Default: claude -p (Get-Content PROMPT.md -Raw) --dangerously-skip-permissions
  • With model: claude -p (Get-Content PROMPT.md -Raw) --model haiku --dangerously-skip-permissions

For Gemini CLI: Replace [AI_COMMAND_WITH_PROMPT] with piped command:

  • Default: Get-Content PROMPT.md -Raw | gemini --yolo
  • With model: Get-Content PROMPT.md -Raw | gemini --model gemini-3-flash --yolo

For Grok CLI: Replace [AI_COMMAND_WITH_PROMPT] with piped command:

  • Default: Get-Content PROMPT.md -Raw | grok-auto
  • With model: Get-Content PROMPT.md -Raw | grok-auto -m grok-4-latest

CRITICAL: Claude Code requires the prompt as a command-line argument. Piping does NOT work with claude -p.

PowerShell - Simple Fixed:

for ($i=1; $i -le N; $i++) { $start = Get-Date; Write-Host "`n=== Run $i/N ===" -ForegroundColor Cyan; [AI_COMMAND_WITH_PROMPT]; $duration = ((Get-Date) - $start).TotalSeconds; Write-Host "⏱️  Time: $([math]::Round($duration, 2))s" -ForegroundColor Magenta }

PowerShell - Simple Fixed + Delay:

for ($i=1; $i -le N; $i++) { $start = Get-Date; Write-Host "`n=== Run $i/N ===" -ForegroundColor Cyan; [AI_COMMAND_WITH_PROMPT]; $duration = ((Get-Date) - $start).TotalSeconds; Write-Host "⏱️  Time: $([math]::Round($duration, 2))s" -ForegroundColor Magenta; if ($i -lt N) { Write-Host "⏸️  Waiting X seconds..." -ForegroundColor Yellow; Start-Sleep -Seconds X } }

PowerShell - Simple Infinite + Delay:

$i=1; while ($true) { $start = Get-Date; Write-Host "`n=== Run $i ===" -ForegroundColor Cyan; [AI_COMMAND_WITH_PROMPT]; $duration = ((Get-Date) - $start).TotalSeconds; Write-Host "⏱️  Time: $([math]::Round($duration, 2))s" -ForegroundColor Magenta; Write-Host "⏸️  Waiting X seconds..." -ForegroundColor Yellow; Start-Sleep -Seconds X; $i++ }

PowerShell - Advanced Full Control:

$end = (Get-Date).AddMinutes(M); for ($i=1; $i -le N -and (Get-Date) -lt $end -and -not (Test-Path STOP.txt); $i++) { $start = Get-Date; Write-Host "`n[$(Get-Date -Format 'HH:mm:ss')] Run $i" -ForegroundColor Cyan; [AI_COMMAND_WITH_PROMPT]; $duration = ((Get-Date) - $start).TotalSeconds; Write-Host "⏱️  Time: $([math]::Round($duration, 2))s" -ForegroundColor Magenta; if ($i -lt N) { Write-Host "⏸️  Waiting X seconds..." -ForegroundColor Yellow; Start-Sleep -Seconds X } }; Write-Host "`n✅ Complete!" -ForegroundColor Green

CMD - Simple Fixed:

for /L %i in (1,1,N) do @(echo. & echo === Run %i === & type PROMPT.md | [AI_COMMAND])

Note: CMD has limited capabilities. For time tracking, recommend PowerShell.

CMD - Simple Infinite + Delay:

for /L %i in (1,0,2) do @(echo. & echo === Run %i === & type PROMPT.md | [AI_COMMAND] & timeout /t X /nobreak > nul)

CMD - Advanced:

for /L %i in (1,1,N) do @(if exist STOP.txt exit & echo. & echo [%time%] Run %i & type PROMPT.md | [AI_COMMAND] & timeout /t X /nobreak > nul)

Note: For time tracking, use PowerShell (see RalphPowerShellComands.md).

Bash - Simple Fixed:

for i in {1..N}; do echo -e "\
=== Run $i/N ==="; start=$(date +%s); cat PROMPT.md | [AI_COMMAND]; dur=$(($(date +%s) - start)); echo "⏱️  Time: ${dur}s"; done

Bash - Simple Fixed + Delay:

for i in {1..N}; do echo -e "\
=== Run $i/N ==="; start=$(date +%s); cat PROMPT.md | [AI_COMMAND]; dur=$(($(date +%s) - start)); echo "⏱️  Time: ${dur}s"; [ $i -lt N ] && { echo "⏸️  Waiting X seconds..."; sleep X; }; done

Bash - Simple Infinite + Delay:

i=1; while :; do echo -e "\
=== Run $i ==="; start=$(date +%s); cat PROMPT.md | [AI_COMMAND]; dur=$(($(date +%s) - start)); echo "⏱️  Time: ${dur}s"; echo "⏸️  Waiting X seconds..."; sleep X; ((i++)); done

Bash - Advanced:

end=$(($(date +%s) + M*60)); for i in {1..N}; do [ $(date +%s) -ge $end ] && break; [ -f STOP.txt ] && break; echo -e "\
[$(date +%H:%M:%S)] Run $i"; start=$(date +%s); cat PROMPT.md | [AI_COMMAND]; dur=$(($(date +%s) - start)); echo "⏱️  Time: ${dur}s"; [ $i -lt N ] && { echo "⏸️  Waiting X seconds..."; sleep X; }; done; echo -e "\
✅ Complete!"

Step 8: Create Command File

Create a timestamped filename in the format: ralphcommand-YYYY-MM-DD-HHMMSS.md

Example: ralphcommand-2026-01-14-233045.md

Use the Write tool to create this file in the current directory with:

File structure:

# Ralph Command

Generated: [timestamp]
Shell: [PowerShell/CMD/Bash]

## Command

[THE ACTUAL COMMAND]


## How to run

1. Make sure you have a PROMPT.md file in this directory
2. Copy the command above
3. Paste into your [PowerShell/CMD/Bash] terminal
4. Press Enter

## How to stop

- Press Ctrl+C at any time
[- OR create STOP.txt: `echo $null > STOP.txt` / `touch STOP.txt`] (if stop file enabled)

## What it does

Runs [claude/gemini/grok] with PROMPT.md as input [N times / for M minutes / until stopped].
[Pauses X seconds between runs.]
[Shows timestamp and run number.]
[Displays execution time for each run.]

Step 9: Notify User

After creating the file, tell the user the exact filename created:

✅ Created ralphcommand-YYYY-MM-DD-HHMMSS.md in your current directory!

The file contains your command ready to run. Just open it and copy the command.

For more variations and explanations, see:
- RalphPowerShellComands.md - Full PowerShell reference
- RalphWindowsCommands.md - Full CMD reference
- RalphLinuxCommands.md - Full Bash reference
- RalphGemini.md - Full Gemini CLI guide
- RalphGrok.md - Full Grok CLI guide

Notes

  • CRITICAL FOR CLAUDE CODE: The prompt MUST be passed as a command-line argument, NOT via stdin pipe. Use claude -p (Get-Content PROMPT.md -Raw) --dangerously-skip-permissions. Piping does NOT work with claude -p.
  • For Gemini and Grok: Stdin piping works fine. Use Get-Content PROMPT.md -Raw | gemini --yolo or Get-Content PROMPT.md -Raw | grok-auto.
  • Model Selection is Optional: Always ask which model, but "Default" option omits the --model flag
  • For Claude Code (PowerShell):

- Default format: claude -p (Get-Content PROMPT.md -Raw) --dangerously-skip-permissions - With model: claude -p (Get-Content PROMPT.md -Raw) --model <model> --dangerously-skip-permissions - Models: haiku, sonnet, opus

  • For Gemini CLI:

- Default format: Get-Content PROMPT.md -Raw | gemini --yolo - With model: Get-Content PROMPT.md -Raw | gemini --model <model> --yolo - Models: gemini-3-flash, gemini-3-pro, gemini-2.5-flash, gemini-2.5-pro - Note: -p flag is deprecated in Gemini

  • For Grok CLI:

- Use grok-auto PowerShell function which calls xAI API directly (perfect for automation) - Default format: Get-Content PROMPT.md -Raw | grok-auto (no -m flag, uses GROK_MODEL env var if set) - With model: Get-Content PROMPT.md -Raw | grok-auto -m <model> - Models: grok-code-fast-1, grok-4-latest, grok-beta, grok-4 - Note: grok-auto is a PowerShell function that calls the xAI API directly (no CLI needed) - Default model (grok-code-fast-1 via GROK_MODEL env var) is recommended for automation loops (fastest responses)

  • Default to "Recommended" options when user is unsure
  • Keep commands on one line when possible for easy copy-paste
  • Always include time tracking - Show how long each run takes using Get-Date (PowerShell) or date +%s (Bash)
  • Always create timestamped file - Format: ralphcommand-YYYY-MM-DD-HHMMSS.md
  • DO NOT include cost tracking - No Get-LastCallCost function, no cost calculations, only time tracking
  • For cost tracking with Claude, tell users to check https://console.anthropic.com for actual API usage
  • When generating the final command for PowerShell:

- For Claude: Replace [AI_COMMAND_WITH_PROMPT] with the full claude command including prompt argument - For Gemini/Grok: The piping is included in the template, just replace [AI_COMMAND] with the command after the pipe

平台分布

OpenClaw

92.79%
按下载量换算16,049

安全审计

VirusTotal

通过

ClawScan

可疑

Static analysis

未展示

权限和风险

执行命令

安装流程涉及命令执行,可能通过 openclaw skills install ralph-loop-writer 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。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