MJHuang666

Agent Relay Auto — Development skill for Claude Code

Development community

How to install Agent Relay Auto

This entry records only its repository, not the path inside it, so there is no exact command to give. Open MJHuang666/agent-relay-auto and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Relay Auto does

A protocol-driven orchestration framework for reliable multi-agent collaboration, context handoff, and automated delivery across Codex, OpenCode, Claude Code, and more.｜面向 Codex、OpenCode、Claude Code 等 Agent 的文件化协议驱动型协作框架，支持可靠的上下文接力与自动化交付。

Alternatives in Development

Claude Scholar — Semi-automated research assistant for academic research and software development 3k ★
Hive — Browser-native hive-mind for CLI coding agents — Claude Code, Codex, Gemini, and OpenCode collaborate as real 473 ★
Pi — PI Wisdom-in-Action Engine v23 — top mandates, scene router, delivery quality gates, anti-patterns, battle mom 107 ★

README

Agent Relay Auto

A protocol-driven orchestration framework for reliable multi-Agent collaboration, durable context handoff, and automated delivery across Codex, OpenCode, Claude Code, and more.

[简体中文](README.zh-CN.md) · [Project introduction](docs/PROJECT_INTRO.en-US.md) · [Full usage guide](docs/AGENT_RELAY_AUTO_USAGE.en-US.md) · [Latest release](https://github.com/MJHuang666/agent-relay-auto/releases/latest)

[![Release](https://img.shields.io/github/v/release/MJHuang666/agent-relay-auto?display_name=tag&color=7C3AED)](https://github.com/MJHuang666/agent-relay-auto/releases/latest) [![License](https://img.shields.io/badge/license-Apache--2.0-0EA5E9)](LICENSE) [![CI](https://img.shields.io/github/actions/workflow/status/MJHuang666/agent-relay-auto/validate-release.yml?label=validation)](https://github.com/MJHuang666/agent-relay-auto/actions)

![Agent Relay Auto workflow](docs/images/agent-relay-auto-workflow.png)

The idea in one sentence

Agent Relay Auto is a lightweight multi-Agent collaboration framework for durable, file-based work across one project repository.

Agent Relay Auto does not ask different Agents to share one chat context. It gives them one durable project state. Requirements, plans, decisions, implementation evidence, review conclusions, handoffs, and final reports live in repository files, so the project can survive a model change, a new Agent, a closed chat window, or a migrated machine.

In automatic mode, the user talks only to the foreground Planner. Runner claims and launches the Implementer and Reviewer in the background; after a valid review, it resumes the exact registered Planner conversation, which summarizes delivery and completes DONE:

User ↔ foreground Planner
          │
          ▼
     requirements + plan
          │
          ▼
background Runner ──► Implementer ──► Reviewer
       ▲                                  │
       └──── automatic repair / re-plan ──┘
                                          │