What happens if an agent-loop-prompt.md already exists?

The skill will detect the existing file and ask for your permission before overwriting it to prevent accidental loss of configuration.

What is the 'Ralph' process?

Ralph is a repeatable task cycle where an autonomous agent studies a plan, picks the next task, implements it, validates the code, updates progress, and commits the changes.

Does this skill require an existing plan?

Yes, the skill will ask you for the path to your implementation plan (e.g., plans/feature.md) to ensure the agent has a roadmap to follow.

What is the purpose of progress.txt?

progress.txt acts as a persistent log where the agent records its completed work, changed files, and any learnings, allowing the loop to maintain state across restarts.

How do I trigger this skill?

You can trigger it by asking Claude to 'set up Ralph', 'create an agent loop', or by using the command /create:agent-loop-prompt.

Autonomous Agent Loop Generator

Name: Autonomous Agent Loop Generator
Author: pruett

bypruett

•

生产力与工作流

Automates the creation of a structured execution cycle for autonomous AI agents to work through implementation plans.

This skill implements the 'Ralph' process, a specialized workflow designed to turn Claude Code into an autonomous development agent. By generating a standardized agent-loop-prompt.md file and initializing a progress tracker, it provides the agent with a repeatable set of instructions to study a plan, implement one task at a time, run validations, and commit changes. This is ideal for developers who want to hand off long-running feature implementations to an AI agent while maintaining strict quality control and progress visibility.

主要功能

01Provides a 'COMPLETE' signal for integration with shell loop scripts

02Includes mandatory validation steps like linting and testing before commits

03Generates a standardized agent-loop-prompt.md file for autonomous cycles

04Initializes progress.txt to track state across multiple iterations

05Enforces a disciplined single-task execution pattern

068 GitHub stars

使用场景

01Automating the implementation of a complex feature from a multi-step markdown plan

02Creating a standardized environment for autonomous AI agents to interact with a codebase

03Setting up a hands-off refactoring workflow that validates and commits incrementally

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add pruett/dotfiles create:agent-loop-prompt

For use in Claude.ai and ChatGPT

主要功能

01Provides a 'COMPLETE' signal for integration with shell loop scripts

02Includes mandatory validation steps like linting and testing before commits

03Generates a standardized agent-loop-prompt.md file for autonomous cycles

04Initializes progress.txt to track state across multiple iterations

05Enforces a disciplined single-task execution pattern

068 GitHub stars

使用场景

01Automating the implementation of a complex feature from a multi-step markdown plan

02Creating a standardized environment for autonomous AI agents to interact with a codebase

03Setting up a hands-off refactoring workflow that validates and commits incrementally

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add pruett/dotfiles create:agent-loop-prompt

For use in Claude.ai and ChatGPT