agentic-coding-flywheel-setup

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Agentic Coding Flywheel Setup (ACFS)

Agentic Coding Flywheel Setup (ACFS)

Skill by ara.so — AI Agent Skills collection
技能来自 ara.so — AI Agent 技能集合

Overview

概述

Agentic Coding Flywheel Setup (ACFS) is a bootstrapping system that transforms a fresh Ubuntu VPS into a professional AI-powered development environment in ~30 minutes. It installs 30+ tools, configures three AI coding agents (Claude Code, Codex CLI, Gemini CLI), and sets up a complete agent coordination stack.
Key Features:
  • One-liner install:
    curl | bash
    to full environment
  • Idempotent: Resume from interruptions automatically
  • Vibe mode: Pre-configured for maximum development velocity
  • Manifest-driven: Single source of truth for all tools and configs
  • Battle-tested stack: 10 Dicklesworthstone tools + 20+ developer utilities
What Gets Installed:
  • Modern shell: zsh + oh-my-zsh + powerlevel10k
  • Language runtimes: bun, uv/Python, Rust, Go
  • AI agents: Claude Code, Codex CLI, Gemini CLI
  • Agent coordination: NTM, MCP Agent Mail, SLB
  • Developer tools: tmux, neovim, ripgrep, fzf, git, gh
  • Cloud CLIs: Vault, Wrangler, Supabase, Vercel
  • Security: checksummed upstream installers, verified downloads
Agentic Coding Flywheel Setup (ACFS) 是一套快速搭建系统,可在约30分钟内将全新Ubuntu VPS转换为专业级AI驱动开发环境。它会安装30余种工具,配置三款AI编码Agent(Claude Code、Codex CLI、Gemini CLI),并搭建完整的Agent协作栈。
核心特性:
  • 一键安装:执行
    curl | bash
    即可完成完整环境搭建
  • 幂等性:可自动从中断处恢复安装
  • Vibe模式:预配置以最大化开发效率
  • 清单驱动:所有工具与配置均有单一可信来源
  • 经实战验证的技术栈:10款Dicklesworthstone工具 + 20余种开发者实用工具
安装内容:
  • 现代Shell:zsh + oh-my-zsh + powerlevel10k
  • 语言运行时:bun、uv/Python、Rust、Go
  • AI Agent:Claude Code、Codex CLI、Gemini CLI
  • Agent协作组件:NTM、MCP Agent Mail、SLB
  • 开发者工具:tmux、neovim、ripgrep、fzf、git、gh
  • 云服务CLI:Vault、Wrangler、Supabase、Vercel
  • 安全保障:带校验和的上游安装程序、已验证的下载源

Installation

安装步骤

Quick Install (Latest)

快速安装(最新版本)

bash
undefined
bash
undefined

Default mode (interactive)

默认模式(交互式)

Vibe mode (non-interactive, passwordless sudo, all dangerous flags enabled)

Vibe模式(非交互式,免密码sudo,启用所有高级功能)

Production Install (Pinned Version)

生产环境安装(固定版本)

bash
undefined
bash
undefined

Pin to a tagged release (recommended)

固定到已标记的发布版本(推荐)

Pin to a specific commit

固定到特定提交版本

undefined
undefined

Installation Flags

安装参数

bash
--yes              # Non-interactive mode (auto-confirm all prompts)
--mode vibe        # Enable vibe mode (passwordless sudo, dangerous flags)
--ref <tag|sha>    # Pin to specific version (tag or commit SHA)
--skip-check       # Skip system requirements check
--debug            # Enable verbose debug output
bash
--yes              # 非交互式模式(自动确认所有提示)
--mode vibe        # 启用Vibe模式(免密码sudo,高级功能)
--ref <tag|sha>    # 固定到特定版本(标签或提交SHA值)
--skip-check       # 跳过系统要求检查
--debug            # 启用详细调试输出

Requirements

系统要求

  • OS: Ubuntu 25.10 (installer auto-upgrades from 24.04/24.10)
  • User: Non-root user with sudo access
  • Network: Internet connection for package downloads
  • Disk: ~10GB free space
  • Memory: 2GB+ RAM recommended
  • 操作系统:Ubuntu 25.10(安装程序可自动从24.04/24.10升级)
  • 用户权限:拥有sudo权限的非root用户
  • 网络:可连接互联网以下载软件包
  • 磁盘空间:约10GB可用空间
  • 内存:推荐2GB及以上内存

Post-Install Commands

安装后命令

After installation, ACFS provides several management commands:
安装完成后,ACFS提供多项管理命令:

Health Check

健康检查

bash
undefined
bash
undefined

Run comprehensive system health check

运行全面系统健康检查

acfs doctor
acfs doctor

Check specific category

检查特定分类

acfs doctor --category shells acfs doctor --category language_runtimes acfs doctor --category agent_clis
undefined
acfs doctor --category shells acfs doctor --category language_runtimes acfs doctor --category agent_clis
undefined

Update ACFS

更新ACFS

bash
undefined
bash
undefined

Update to latest version

更新到最新版本

acfs update
acfs update

Update to specific version

更新到特定版本

acfs update --ref v0.7.0
undefined
acfs update --ref v0.7.0
undefined

Service Setup

服务配置

bash
undefined
bash
undefined

Configure agent coordination services (NTM, MCP Agent Mail, etc.)

配置Agent协作服务(NTM、MCP Agent Mail等)

acfs services-setup
undefined
acfs services-setup
undefined

Onboarding

新手引导

bash
undefined
bash
undefined

Interactive tutorial for new users

面向新用户的交互式教程

onboard
undefined
onboard
undefined

Key Tools & Usage

核心工具及使用方法

AI Coding Agents

AI编码Agent

Claude Code (Anthropic)
bash
undefined
Claude Code(Anthropic)
bash
undefined

Set API key

设置API密钥

export ANTHROPIC_API_KEY="sk-ant-..."
export ANTHROPIC_API_KEY="sk-ant-..."

Start interactive session

启动交互式会话

claude
claude

One-shot command

单次命令执行

claude "refactor this function to use async/await" < input.js
claude "将此函数重构为使用async/await" < input.js

Edit files directly

直接编辑文件

claude --edit main.py "add error handling"

**Codex CLI** (OpenAI)
```bash
claude --edit main.py "添加错误处理"

**Codex CLI**(OpenAI)
```bash

Set API key

设置API密钥

export OPENAI_API_KEY="sk-..."
export OPENAI_API_KEY="sk-..."

Interactive mode

交互式模式

codex
codex

Generate code

生成代码

codex "write a REST API with Express"
codex "使用Express编写一个REST API"

Code review

代码审查

codex --review src/

**Gemini CLI** (Google)
```bash
codex --review src/

**Gemini CLI**(Google)
```bash

Set API key

设置API密钥

export GEMINI_API_KEY="..."
export GEMINI_API_KEY="..."

Start session

启动会话

gemini
gemini

Multi-modal (with images)

多模态(支持图片)

gemini "analyze this screenshot" --image screenshot.png
undefined
gemini "分析此截图" --image screenshot.png
undefined

Agent Coordination Stack

Agent协作栈

NTM (Note-Taking Manager)
bash
undefined
NTM(笔记管理工具)
bash
undefined

Initialize notes directory

初始化笔记目录

ntm init
ntm init

Create note

创建新笔记

ntm new "project-ideas"
ntm new "项目创意"

Search notes

搜索笔记

ntm search "api design"
ntm search "API设计"

List all notes

列出所有笔记

ntm list
ntm list

Edit note

编辑笔记

ntm edit "project-ideas"

**MCP Agent Mail** (inter-agent communication)
```bash
ntm edit "项目创意"

**MCP Agent Mail**(Agent间通信工具)
```bash

Start mail server

启动邮件服务器

mcp_agent_mail serve --port 3000
mcp_agent_mail serve --port 3000

Send message between agents

在Agent间发送消息

mcp_agent_mail send --to claude@agents.local --subject "Review PR" --body "..."
mcp_agent_mail send --to claude@agents.local --subject "审查PR" --body "..."

Check inbox

查看收件箱

mcp_agent_mail inbox

**SLB (Session Ledger Bot)**
```bash
mcp_agent_mail inbox

**SLB(会话记录Bot)**
```bash

Initialize session tracking

初始化会话跟踪

slb init
slb init

Start new session

启动新会话

slb start "implementing auth feature"
slb start "实现认证功能"

Log activity

记录活动

slb log "completed OAuth flow"
slb log "完成OAuth流程"

End session

结束会话

slb end
slb end

View session history

查看会话历史

slb history

**BV (Beads Viewer)** - Session visualization
```bash
slb history

**BV(会话可视化工具)**
```bash

View current session beads

查看当前会话数据

bv show
bv show

Export session graph

导出会话图表

bv export --format svg --output session.svg
bv export --format svg --output session.svg

Analyze session patterns

分析会话模式

bv analyze

**UBS (Universal Backup System)**
```bash
bv analyze

**UBS(通用备份系统)**
```bash

Configure backup

配置备份

ubs init --dest s3://my-backups
ubs init --dest s3://my-backups

Create backup

创建备份

ubs backup ~/projects
ubs backup ~/projects

List backups

列出备份

ubs list
ubs list

Restore backup

恢复备份

ubs restore --backup 2025-01-15 --dest ~/restored
undefined
ubs restore --backup 2025-01-15 --dest ~/restored
undefined

Developer Tools

开发者工具

Tmux (terminal multiplexer)
bash
undefined
Tmux(终端复用工具)
bash
undefined

ACFS includes pre-configured tmux.conf

ACFS包含预配置的tmux.conf

Start new session

启动新会话

tmux new -s work
tmux new -s work

Detach: Ctrl+b d

分离会话:Ctrl+b d

List sessions

列出会话

tmux ls
tmux ls

Attach to session

连接到会话

tmux attach -t work
tmux attach -t work

Split panes

拆分窗格

Horizontal: Ctrl+b "

水平拆分:Ctrl+b "

Vertical: Ctrl+b %

垂直拆分:Ctrl+b %


**Neovim** (text editor)
```bash

**Neovim**(文本编辑器)
```bash

Launch neovim

启动neovim

nvim file.py
nvim file.py

ACFS includes sensible defaults

ACFS包含合理的默认配置

Basic commands:

基础命令:

:w - save

:w - 保存

:q - quit

:q - 退出

:wq - save and quit

:wq - 保存并退出

i - insert mode

i - 插入模式

ESC - normal mode

ESC - 普通模式


**Ripgrep** (fast search)
```bash

**Ripgrep**(快速搜索工具)
```bash

Search current directory

搜索当前目录

rg "function.*async"
rg "function.*async"

Search with context

带上下文的搜索

rg -C 3 "TODO"
rg -C 3 "TODO"

Search specific file types

搜索特定文件类型

rg --type py "class.*Model"
rg --type py "class.*Model"

Case-insensitive

不区分大小写

rg -i "error"

**FZF** (fuzzy finder)
```bash
rg -i "error"

**FZF**(模糊查找工具)
```bash

Interactive file finder (Ctrl+T in shell)

交互式文件查找(Shell中按Ctrl+T)

Interactive history search (Ctrl+R in shell)

交互式历史搜索(Shell中按Ctrl+R)

In scripts

在脚本中使用

selected_file=$(fzf) echo "Selected: $selected_file"
selected_file=$(fzf) echo "已选择: $selected_file"

Filter command output

过滤命令输出

ps aux | fzf
undefined
ps aux | fzf
undefined

Configuration

配置说明

ACFS Config Files

ACFS配置文件

All ACFS configuration lives in
~/.acfs/
:
bash
~/.acfs/
├── zshrc              # Shell configuration
├── tmux.conf          # Tmux configuration
├── state.json         # Installation state
├── onboard/           # Tutorial lessons
└── scripts/           # Utility scripts
所有ACFS配置均位于
~/.acfs/
目录下:
bash
~/.acfs/
├── zshrc              # Shell配置
├── tmux.conf          # Tmux配置
├── state.json         # 安装状态
├── onboard/           # 教程内容
└── scripts/           # 实用脚本

Environment Variables

环境变量

Key environment variables to configure:
bash
undefined
需配置的核心环境变量:
bash
undefined

AI Agent API Keys

AI Agent API密钥

export ANTHROPIC_API_KEY="sk-ant-..." export OPENAI_API_KEY="sk-..." export GEMINI_API_KEY="..."
export ANTHROPIC_API_KEY="sk-ant-..." export OPENAI_API_KEY="sk-..." export GEMINI_API_KEY="..."

Cloud Provider Keys

云服务商密钥

export CLOUDFLARE_API_TOKEN="..." export SUPABASE_ACCESS_TOKEN="..." export VERCEL_TOKEN="..."
export CLOUDFLARE_API_TOKEN="..." export SUPABASE_ACCESS_TOKEN="..." export VERCEL_TOKEN="..."

Vault (secrets management)

Vault(密钥管理)

export VAULT_ADDR="https://vault.example.com" export VAULT_TOKEN="..."
export VAULT_ADDR="https://vault.example.com" export VAULT_TOKEN="..."

ACFS behavior

ACFS行为配置

export ACFS_MODE="vibe" # Enable vibe mode features export ACFS_AUTO_UPDATE="true" # Auto-update ACFS export ACFS_TELEMETRY="false" # Disable telemetry

Add to `~/.zshrc`:
```bash
echo 'export ANTHROPIC_API_KEY="sk-ant-..."' >> ~/.zshrc
source ~/.zshrc
export ACFS_MODE="vibe" # 启用Vibe模式特性 export ACFS_AUTO_UPDATE="true" # 自动更新ACFS export ACFS_TELEMETRY="false" # 禁用遥测

添加到 `~/.zshrc`:
```bash
echo 'export ANTHROPIC_API_KEY="sk-ant-..."' >> ~/.zshrc
source ~/.zshrc

Vibe Mode

Vibe模式

Vibe mode enables maximum development velocity:
bash
undefined
Vibe模式可最大化开发效率:
bash
undefined

Enable during install

安装时启用

curl -fsSL "..." | bash -s -- --mode vibe
curl -fsSL "..." | bash -s -- --mode vibe

Features enabled in vibe mode:

Vibe模式启用的特性:

- Passwordless sudo for your user

- 当前用户免密码sudo

- All dangerous agent flags enabled

- 启用所有Agent高级功能

- Auto-confirm for tool installations

- 自动确认工具安装

- Optimized shell aliases

- 优化后的Shell别名

- Pre-configured git settings

- 预配置的Git设置


**Security Note**: Vibe mode is designed for personal development VPS instances, not production servers or shared environments.

**安全提示**:Vibe模式专为个人开发VPS实例设计,请勿在生产服务器或共享环境中使用。

Common Patterns

常见使用场景

Pattern 1: Initial VPS Setup

场景1:初始VPS配置

bash
undefined
bash
undefined

1. SSH into fresh Ubuntu VPS

1. SSH连接到全新Ubuntu VPS

ssh user@vps-ip
ssh user@vps-ip

2. Run ACFS installer

2. 运行ACFS安装程序

3. Set up API keys

3. 设置API密钥

echo 'export ANTHROPIC_API_KEY="sk-ant-..."' >> ~/.zshrc echo 'export OPENAI_API_KEY="sk-..."' >> ~/.zshrc source ~/.zshrc
echo 'export ANTHROPIC_API_KEY="sk-ant-..."' >> ~/.zshrc echo 'export OPENAI_API_KEY="sk-..."' >> ~/.zshrc source ~/.zshrc

4. Verify installation

4. 验证安装

acfs doctor
acfs doctor

5. Start coding with agents

5. 使用Agent开始编码

claude "create a new Express API project"
undefined
claude "创建一个新的Express API项目"
undefined

Pattern 2: Agent Session Workflow

场景2:Agent会话工作流

bash
undefined
bash
undefined

Start session tracking

启动会话跟踪

slb start "building user authentication"
slb start "构建用户认证系统"

Use AI agent for scaffolding

使用AI Agent搭建基础代码

claude "create user model with email and password fields"
claude "创建包含邮箱和密码字段的用户模型"

Track progress

记录进度

slb log "created user model"
slb log "已创建用户模型"

Review with another agent

使用另一Agent进行代码审查

codex --review models/user.py
codex --review models/user.py

Visualize session

可视化会话

bv show
bv show

End session

结束会话

slb end
undefined
slb end
undefined

Pattern 3: Multi-Agent Collaboration

场景3:多Agent协作

bash
undefined
bash
undefined

Agent 1: Generate implementation

Agent 1:生成实现代码

claude "implement JWT authentication" > auth.py
claude "实现JWT认证" > auth.py

Agent 2: Review for security

Agent 2:进行安全审查

codex --review auth.py > review.md
codex --review auth.py > review.md

Agent 3: Optimize performance

Agent 3:优化性能

gemini "optimize this authentication flow" < auth.py > auth_optimized.py
gemini "优化此认证流程" < auth.py > auth_optimized.py

Track collaboration via MCP Agent Mail

通过MCP Agent Mail跟踪协作

mcp_agent_mail send --to review-agent --body "$(cat review.md)"
undefined
mcp_agent_mail send --to review-agent --body "$(cat review.md)"
undefined

Pattern 4: Backup Before Experimentation

场景4:实验前备份

bash
undefined
bash
undefined

Backup current state

备份当前状态

ubs backup ~/projects/myapp
ubs backup ~/projects/myapp

Let agent make risky changes

让Agent进行高风险修改

claude --edit src/ "refactor entire codebase to use TypeScript" --allow-dangerous
claude --edit src/ "将整个代码库重构为TypeScript" --allow-dangerous

If things go wrong

如果出现问题,恢复备份

ubs restore --backup latest --dest ~/projects/myapp
undefined
ubs restore --backup latest --dest ~/projects/myapp
undefined

Pattern 5: Reproducible Team Environments

场景5:可复现的团队环境

bash
undefined
bash
undefined

Team lead: Create bootstrap script

团队负责人:创建引导脚本

cat > setup-team-env.sh << 'EOF' #!/bin/bash
cat > setup-team-env.sh << 'EOF' #!/bin/bash

Install ACFS with pinned version

安装固定版本的ACFS

Configure team-specific tools

配置团队专属工具

git config --global user.name "$GIT_NAME" git config --global user.email "$GIT_EMAIL"
git config --global user.name "$GIT_NAME" git config --global user.email "$GIT_EMAIL"

Clone team repos

克隆团队仓库

gh repo clone myorg/backend ~/projects/backend gh repo clone myorg/frontend ~/projects/frontend EOF
gh repo clone myorg/backend ~/projects/backend gh repo clone myorg/frontend ~/projects/frontend EOF

Team members: Run identical setup

团队成员:运行相同的配置脚本

export GIT_NAME="Alice" GIT_EMAIL="alice@company.com" bash setup-team-env.sh
undefined
export GIT_NAME="Alice" GIT_EMAIL="alice@company.com" bash setup-team-env.sh
undefined

Architecture

架构设计

Manifest-Driven Design

清单驱动设计

ACFS uses a single source of truth:
acfs.manifest.yaml
yaml
undefined
ACFS使用单一可信来源:
acfs.manifest.yaml
yaml
undefined

Example manifest structure (simplified)

示例清单结构(简化版)

modules:
  • id: bun category: language_runtimes install: bash: | curl -fsSL https://bun.sh/install | bash verify: bash: | command -v bun >/dev/null 2>&1
  • id: claude_code category: agent_clis install: bash: | npm install -g @anthropic-ai/claude-code verify: bash: | command -v claude >/dev/null 2>&1
undefined
modules:
  • id: bun category: language_runtimes install: bash: | curl -fsSL https://bun.sh/install | bash verify: bash: | command -v bun >/dev/null 2>&1
  • id: claude_code category: agent_clis install: bash: | npm install -g @anthropic-ai/claude-code verify: bash: | command -v claude >/dev/null 2>&1
undefined

Generated Scripts (Reference)

生成脚本(参考)

The manifest generates reference scripts in
scripts/generated/
:
bash
scripts/generated/
├── install_shells.sh
├── install_language_runtimes.sh
├── install_agent_clis.sh
├── install_agent_coordination.sh
├── install_developer_tools.sh
├── install_cloud_clis.sh
├── doctor_checks.sh
└── install_all.sh
清单会在
scripts/generated/
目录下生成参考脚本:
bash
scripts/generated/
├── install_shells.sh
├── install_language_runtimes.sh
├── install_agent_clis.sh
├── install_agent_coordination.sh
├── install_developer_tools.sh
├── install_cloud_clis.sh
├── doctor_checks.sh
└── install_all.sh

Security

安全机制

ACFS implements multiple security layers:
  1. Checksummed installers:
    checksums.yaml
    verifies upstream scripts
  2. HTTPS-only downloads: All sources fetched over TLS
  3. Isolated environments: Each tool runs in its own context
  4. API key isolation: Keys stored in environment, not config files
  5. Audit logging: All installations logged to
    ~/.acfs/state.json
ACFS实现了多层安全保障:
  1. 带校验和的安装程序
    checksums.yaml
    验证上游脚本完整性
  2. 仅HTTPS下载:所有资源均通过TLS协议获取
  3. 隔离环境:每个工具在独立环境中运行
  4. API密钥隔离:密钥存储在环境变量中,而非配置文件
  5. 审计日志:所有安装操作均记录到
    ~/.acfs/state.json

Troubleshooting

故障排查

Installation Fails Midway

安装中途失败

bash
undefined
bash
undefined

ACFS is idempotent - just re-run

ACFS具有幂等性,只需重新运行安装命令

Check installation state

检查安装状态

cat ~/.acfs/state.json | jq '.completed_phases'
cat ~/.acfs/state.json | jq '.completed_phases'

Run doctor to identify issues

运行健康检查工具定位问题

acfs doctor
undefined
acfs doctor
undefined

Doctor Check Failures

健康检查失败

bash
undefined
bash
undefined

Run full diagnostic

运行完整诊断

acfs doctor --verbose
acfs doctor --verbose

Fix common issues:

修复常见问题:

- Missing PATH entries

- 缺失PATH条目

echo 'export PATH="$HOME/.bun/bin:$PATH"' >> ~/.zshrc source ~/.zshrc
echo 'export PATH="$HOME/.bun/bin:$PATH"' >> ~/.zshrc source ~/.zshrc

- Incomplete installations

- 安装不完整

acfs update
acfs update

- Permission issues

- 权限问题

sudo chown -R $USER:$USER ~/.acfs
undefined
sudo chown -R $USER:$USER ~/.acfs
undefined

Agent CLI Not Found

Agent CLI未找到

bash
undefined
bash
undefined

Verify installation

验证安装状态

acfs doctor --category agent_clis
acfs doctor --category agent_clis

Reinstall specific agent

重新安装特定Agent

(Manually, as ACFS doesn't have per-tool reinstall yet)

npm install -g @anthropic-ai/claude-code
#(目前ACFS暂不支持单工具重装,需手动执行) npm install -g @anthropic-ai/claude-code

Check API key

检查API密钥

echo $ANTHROPIC_API_KEY # Should not be empty
undefined
echo $ANTHROPIC_API_KEY # 不应为空
undefined

Slow Performance

性能缓慢

bash
undefined
bash
undefined

Check system resources

检查系统资源

htop
htop

Verify zsh compilation

验证zsh编译

rm ~/.zcompdump zsh -c 'autoload -U compinit && compinit'
rm ~/.zcompdump zsh -c 'autoload -U compinit && compinit'

Disable plugins if needed

如有需要,禁用部分插件

Edit ~/.acfs/zshrc and comment out heavy plugins

编辑 ~/.acfs/zshrc 并注释掉资源占用高的插件

undefined
undefined

SSH Connection Issues After Install

安装后SSH连接问题

bash
undefined
bash
undefined

If you set up SSH key during install but can't connect:

如果安装时设置了SSH密钥但无法连接:

From local machine - use password one more time

从本地机器使用密码再次连接

ssh user@vps-ip
ssh user@vps-ip

On VPS - verify SSH key was added

在VPS上验证SSH密钥是否已添加

cat ~/.ssh/authorized_keys
cat ~/.ssh/authorized_keys

If missing, add your public key

如果缺失,添加你的公钥

echo "ssh-ed25519 AAAA..." >> ~/.ssh/authorized_keys chmod 600 ~/.ssh/authorized_keys
echo "ssh-ed25519 AAAA..." >> ~/.ssh/authorized_keys chmod 600 ~/.ssh/authorized_keys

Try key-based auth again from local

从本地机器再次尝试密钥认证

ssh -i ~/.ssh/id_ed25519 user@vps-ip
undefined
ssh -i ~/.ssh/id_ed25519 user@vps-ip
undefined

Services Won't Start

服务无法启动

bash
undefined
bash
undefined

Check service status

检查服务状态

acfs services-setup --status
acfs services-setup --status

View logs

查看日志

journalctl -u ntm -f journalctl -u mcp_agent_mail -f
journalctl -u ntm -f journalctl -u mcp_agent_mail -f

Restart services

重启服务

sudo systemctl restart ntm sudo systemctl restart mcp_agent_mail
undefined
sudo systemctl restart ntm sudo systemctl restart mcp_agent_mail
undefined

Advanced Usage

高级用法

Custom Tool Installation

自定义工具安装

Add custom tools to your environment post-install:
bash
undefined
安装完成后可添加自定义工具到环境中:
bash
undefined

Install additional language runtime

安装额外的语言运行时

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

Install additional CLI tool

安装额外的CLI工具

cargo install ripgrep
cargo install ripgrep

Update PATH if needed

如有需要,更新PATH

echo 'export PATH="$HOME/.cargo/bin:$PATH"' >> ~/.zshrc source ~/.zshrc
echo 'export PATH="$HOME/.cargo/bin:$PATH"' >> ~/.zshrc source ~/.zshrc

Verify with doctor

使用健康检查工具验证

acfs doctor
undefined
acfs doctor
undefined

Extending the Manifest

扩展清单

To add tools to ACFS itself, edit
acfs.manifest.yaml
:
yaml
modules:
  - id: my_custom_tool
    category: developer_tools
    description: "My custom development tool"
    install:
      bash: |
        curl -fsSL https://example.com/install.sh | bash
    verify:
      bash: |
        command -v my_tool >/dev/null 2>&1
Then regenerate scripts:
bash
cd packages/manifest
npm run generate
如需向ACFS添加工具,编辑
acfs.manifest.yaml
yaml
modules:
  - id: my_custom_tool
    category: developer_tools
    description: "我的自定义开发工具"
    install:
      bash: |
        curl -fsSL https://example.com/install.sh | bash
    verify:
      bash: |
        command -v my_tool >/dev/null 2>&1
然后重新生成脚本:
bash
cd packages/manifest
npm run generate

Integration Testing

集成测试

ACFS includes Docker-based integration tests:
bash
undefined
ACFS包含基于Docker的集成测试:
bash
undefined

Run full install test in Docker

在Docker中运行完整安装测试

bash tests/vm/test_install_ubuntu.sh
bash tests/vm/test_install_ubuntu.sh

Test specific version

测试特定版本

ACFS_REF=v0.7.0 bash tests/vm/test_install_ubuntu.sh
ACFS_REF=v0.7.0 bash tests/vm/test_install_ubuntu.sh

Test vibe mode

测试Vibe模式

ACFS_MODE=vibe bash tests/vm/test_install_ubuntu.sh
undefined
ACFS_MODE=vibe bash tests/vm/test_install_ubuntu.sh
undefined

Monitoring Agent Activity

监控Agent活动

bash
undefined
bash
undefined

Track all agent sessions

跟踪所有Agent会话

slb history --all
slb history --all

Export session data

导出会话数据

slb export --format json > sessions.json
slb export --format json > sessions.json

Visualize patterns

可视化会话模式

bv analyze --input sessions.json --output report.html
bv analyze --input sessions.json --output report.html

Monitor in real-time

实时监控

tail -f ~/.acfs/logs/agent-activity.log
undefined
tail -f ~/.acfs/logs/agent-activity.log
undefined

Version Compatibility

版本兼容性

  • v0.7.0: Current stable (Ubuntu 25.10, all features)
  • v0.6.x: Legacy (Ubuntu 24.10, limited stack tools)
  • v0.5.x: Deprecated (Ubuntu 24.04, no vibe mode)
Always use the latest tagged release for production environments.
  • v0.7.0:当前稳定版(支持Ubuntu 25.10,包含所有功能)
  • v0.6.x:旧版本(支持Ubuntu 24.10,工具栈有限)
  • v0.5.x:已弃用(支持Ubuntu 24.04,无Vibe模式)
生产环境请始终使用最新的标记发布版本。

Additional Resources

额外资源