huashu-agent-swarm

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Infinite Agent Loop - 无限Agent蜂群模式

Infinite Agent Loop - Infinite Agent Swarm Mode

受Nicholas Carlini用16个Claude实例自主构建C编译器的启发。没有master agent，纯git自组织，每个agent独立认领任务、写代码、推送。

Inspired by Nicholas Carlini using 16 Claude instances to independently build a C compiler. No master agent, pure Git self-organization, each agent independently claims tasks, writes code, and pushes.

触发条件

Trigger Conditions

当用户提到「蜂群模式」「多agent并行」「infinite loop」「agent swarm」「启动蜂群」时使用此技能。

Use this skill when users mention "swarm mode", "multi-agent parallel", "infinite loop", "agent swarm", "start swarm".

前置要求

Prerequisites

tmux（
```
brew install tmux
```
）
claude CLI（已安装）
git 仓库（已有或新建）

tmux (
```
brew install tmux
```
)
claude CLI (installed)
Git repository (existing or newly created)

使用流程

Usage Process

Step 1: 描述项目

Step 1: Describe the Project

用户告诉我：

项目目录路径（必须是git仓库）
项目目标和总体描述
初始任务列表（或让agent自行拆解）
agent数量（默认8个）
代码规范和测试命令

Users tell me:

Project directory path (must be a Git repository)
Project goals and overall description
Initial task list (or let agents break down tasks automatically)
Number of agents (default: 8)
Code specifications and test commands

Step 2: 初始化项目

Step 2: Initialize the Project

bash

bash SKILL_DIR/scripts/setup_project.sh <项目目录>

这会在项目中创建：

```
AGENT_PROMPT.md
```
- 从模板生成，需要我根据用户需求定制
```
TASKS.md
```
- 初始任务清单
```
current_tasks/
```
- 任务认领目录
```
agent_logs/
```
- 日志目录

然后我根据

references/agent-prompt-template.md

定制

AGENT_PROMPT.md

，填入项目具体信息。

bash

bash SKILL_DIR/scripts/setup_project.sh <project directory>

This will create in the project:

```
AGENT_PROMPT.md
```
- Generated from template, needs to be customized according to user requirements
```
TASKS.md
```
- Initial task list
```
current_tasks/
```
- Task claim directory
```
agent_logs/
```
- Log directory

Then I customize

AGENT_PROMPT.md

based on

references/agent-prompt-template.md

, filling in specific project information.

Step 3: 启动蜂群

Step 3: Start the Swarm

bash

bash SKILL_DIR/scripts/start_swarm.sh <agent数量> <项目目录>

这会：

为每个agent创建 git worktree（共享.git对象库，不浪费磁盘）
创建 tmux session，每个pane一个agent
每个agent进入无限循环：pull → 认领任务 → 执行 → push → 下一个

bash

bash SKILL_DIR/scripts/start_swarm.sh <number of agents> <project directory>

This will:

Create a git worktree for each agent (shares .git object library, no disk waste)
Create a tmux session, one agent per pane
Each agent enters an infinite loop: pull → claim task → execute → push → next

Step 4: 打开观测台

Step 4: Open the Dashboard

bash

python3 SKILL_DIR/scripts/dashboard.py <项目目录> 8420

浏览器打开 http://localhost:8420，可以：

实时查看所有agent状态、git log、任务进度
查看每个agent的最新日志
输入框直接发送指令给agent（写入HUMAN_INPUT.md）
一键停止所有agent

也可以用命令行监控：

bash

undefined

bash

python3 SKILL_DIR/scripts/dashboard.py <project directory> 8420

Open http://localhost:8420 in browser, you can:

View real-time status of all agents, git log, task progress
View the latest logs of each agent
Send instructions directly to agents via input box (writes to HUMAN_INPUT.md)
Stop all agents with one click

You can also monitor via command line:

bash

undefined

终端状态

Terminal status

bash SKILL_DIR/scripts/status.sh <项目目录>

bash SKILL_DIR/scripts/status.sh <project directory>

发送指令

Send instructions

bash SKILL_DIR/scripts/send_input.sh <项目目录> "你的指令"

bash SKILL_DIR/scripts/send_input.sh <project directory> "your instruction"

直接进入tmux观察

Directly enter tmux to observe

tmux attach -t swarm-<项目名>

undefined

tmux attach -t swarm-<project name>

undefined

Step 5: 停止

Step 5: Stop

bash

bash SKILL_DIR/scripts/stop_swarm.sh <项目目录>

自动停止所有agent + 合并分支 + 清理worktrees。

bash

bash SKILL_DIR/scripts/stop_swarm.sh <project directory>

Automatically stops all agents + merges branches + cleans up worktrees.

核心机制

Core Mechanisms

Git自组织协调

Git Self-Organization Coordination

每个agent通过
```
current_tasks/*.lock
```
文件认领任务
通过
```
TASKS.md
```
了解全局进度
通过
```
git log
```
了解其他agent的工作
冲突由agent自行解决

Each agent claims tasks via
```
current_tasks/*.lock
```
files
Understands global progress through
```
TASKS.md
```
Learns about other agents' work through
```
git log
```
Conflicts are resolved by agents themselves

Git Worktree隔离

Git Worktree Isolation

不用多份clone，用
```
git worktree
```
实现隔离
所有worktree共享同一个
```
.git
```
对象库
每个agent在自己的worktree独立工作

No multiple clones, uses
```
git worktree
```
for isolation
All worktrees share the same
```
.git
```
object library
Each agent works independently in its own worktree

无限循环

Infinite Loop

每个agent完成一个session后自动开始下一个
通过
```
git pull
```
获取其他agent的最新成果
通过 sleep 间隔避免API限流

Each agent automatically starts the next session after completing one
Gets the latest results from other agents via
```
git pull
```
Uses sleep intervals to avoid API rate limits

关键配置

Key Configurations

参数	默认值	说明
agent数量	8	可在启动时指定
sleep间隔	5秒	agent_loop.sh中可调
模型	claude-opus-4-6	agent_loop.sh中可调

Parameter	Default Value	Description
Number of agents	8	Specifiable at startup
Sleep interval	5 seconds	Adjustable in agent_loop.sh
Model	claude-opus-4-6	Adjustable in agent_loop.sh

风险和应对

Risks and Countermeasures

风险	应对
API限流	sleep间隔 + 可调agent数量
合并冲突	AGENT_PROMPT指导小粒度commit
死循环无用功	日志监控 + 停止条件
磁盘空间	stop_swarm.sh自动清理
成本失控	可在AGENT_PROMPT中限制session数

花叔出品 | AI Native Coder · 独立开发者公众号「花叔」| 30万+粉丝 | AI工具与效率提升代表作：小猫补光灯（AppStore付费榜Top1）·《一本书玩转DeepSeek》

Risk	Countermeasure
API rate limits	Sleep intervals + adjustable number of agents
Merge conflicts	AGENT_PROMPT guides small-grained commits
Infinite loop of useless work	Log monitoring + stop conditions
Disk space issues	stop_swarm.sh automatically cleans up
Uncontrolled costs	Limit number of sessions in AGENT_PROMPT

Produced by Uncle Hua | AI Native Coder · Independent Developer Official Account "Uncle Hua" | 300,000+ followers | AI Tools & Efficiency Improvement Masterpieces: Kitten Fill Light (Top 1 in AppStore Paid Charts) · "Mastering DeepSeek in One Book"