What is the Jin Ralph skill?

Jin Ralph is a specialized workflow for Claude Code that creates a self-referencing loop to execute tasks and verify them independently until the task is successfully completed.

What happens if a task is too complex for the standard agent?

Jin Ralph features complexity escalation; if the standard swe-agent fails twice, the system automatically switches to the more powerful swe-agent-high (powered by Opus) for the remaining attempts.

How many times will Jin Ralph attempt to fix a problem?

The skill is designed to run the execution and verification loop up to 5 times. If it cannot succeed within 5 attempts, it generates a final report for manual review.

How does the verification process avoid bias?

The verification agent (swe-verifier) only receives the original request and the resulting code diff, ensuring it judges the outcome objectively without being influenced by the implementation agent's reasoning.

Jin Ralph Self-Correction Loop

Name: Jin Ralph Self-Correction Loop
Author: cjinzy

bycjinzy

•

开发者工具

Orchestrates an autonomous execute-verify-modify loop that iteratively refines code until it passes independent quality verification.

Jin Ralph is a sophisticated self-referencing loop designed for Claude Code to ensure high-reliability task completion. It operates by delegating implementation to specialized SWE agents and then subjecting the results to an independent verifier. If the verifier identifies issues, the skill automatically incorporates the feedback into a new execution cycle. With support for up to five iterations and automatic escalation from Sonnet to Opus for complex challenges, Jin Ralph minimizes human intervention by autonomously converging on high-quality solutions through rigorous feedback loops.

主要功能

01Feedback-driven modification cycles that target specific verifier concerns

02Comprehensive Ralph Reports detailing iteration history and failure patterns

03Automatic complexity escalation to Claude 3 Opus after repeated failures

042 GitHub stars

05Autonomous 'Execute-Verify-Modify' loop for iterative task refinement

06Independent verification system to prevent implementation bias and ensure quality

使用场景

01Automating complex bug fixes that require iterative testing and refinement

02Implementing cross-module features where consistency and regression checks are vital

03Refactoring legacy code with autonomous validation to ensure no breaking changes

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cjinzy/jin-claude jin-ralph

For use in Claude.ai and ChatGPT