majestic-marketplace
💡 摘要
Majestic Marketplace 提供一个协作平台,让工程师在编码工作流程中协调 AI 代理。
🎯 适合人群
🤖 AI 吐槽: “看起来很能打,但别让配置把人劝退。”
风险:Medium。建议检查:是否执行 shell/命令行指令;是否发起外网请求(SSRF/数据外发);文件读写范围与路径穿越风险;依赖锁定与供应链风险。以最小权限运行,并在生产环境启用前审计代码与依赖。
Majestic Marketplace
The Majestic marketplace where we share our workflows.
New here? Check out the Marketplace Tutorial for an interactive walkthrough.
The Shift: Engineers Orchestrate Agents
Coding is no longer the bottleneck. Planning, review, and learning loops matter more than syntax. Each feature makes the next one easier to build.
Plan → Work → Assess → Reflect
| Step | What Happens | Key Tools |
|------|--------------|-----------|
| Plan | Agents research codebase + best practices, produce detailed implementation plans | /majestic:plan, architect agent |
| Work | Agents write code, tests, and iterate using real app feedback | /majestic:build-task, coder skills |
| Assess | Multi-angle review: security, performance, simplicity, conventions | /majestic:code-review, quality-gate |
| Reflect | Analyze session patterns, capture lessons so future agents improve | /majestic-tools:insight:reflect, /majestic:add-lesson |
See the Workflow Guide for detailed documentation.
Quick Start
Run the installer:
curl -fsSL https://raw.githubusercontent.com/majesticlabs-dev/majestic-marketplace/master/install.sh | bash
This gives you options to:
- Add marketplace (enables plugin installation)
- Add output styles (formatting guides)
- Add MCP servers (Sequential Thinking)
- Configure shell settings (env vars + alias)
- Install Beads (AI agent memory)
- Install all
Manual Installation
Claude Code
Run claude and add the marketplace:
/plugin marketplace add https://github.com/majesticlabs-dev/majestic-marketplace.git
Then install a plugin:
/plugin install {plugin-name}
Codex CLI (OpenAI)
Export Majestic skills and commands to Codex CLI:
# Interactive selection ./scripts/install-codex.sh # Install all plugins ./scripts/install-codex.sh --all # Install specific plugins ./scripts/install-codex.sh engineer rails tools
This copies skills to ~/.codex/skills/ and commands to ~/.codex/prompts/.
Limitations: Codex doesn't support Claude Code's Task tool, so subagents (parallel research) won't run. Core skills and commands still provide valuable guidance.
Majestic Plugins
| Plugin | Description | |--------|-------------| | majestic-engineer | Language-agnostic engineering workflows (17 agents, 8 commands, 12 skills) | | majestic-rails | Ruby on Rails development tools (23 agents, 5 commands, 9 skills) | | majestic-react | Modern React development with TypeScript (3 agents, 1 command, 4 skills) | | majestic-python | Python development tools (2 agents) | | majestic-devops | Infrastructure-as-Code and DevOps workflows (1 agent, 8 skills) | | majestic-llm | External LLM integration - Codex, Gemini (5 agents, 1 command) | | majestic-marketing | Marketing and SEO tools (14 agents, 2 commands, 24 skills) | | majestic-sales | Sales acceleration tools (1 command, 6 skills) | | majestic-company | Business operations and CEO tools (2 agents, 21 skills) | | majestic-experts | Expert panel discussion system (2 agents, 1 command) | | majestic-tools | Claude Code customization tools (10 commands, 3 skills) |
Recommended Tools
Anthropic Official
| Plugin | Description | Install |
|--------|-------------|---------|
| ralph-loop | Continuous AI agent loop - iterates until task completion using stop hooks | /plugin marketplace add anthropics/claude-plugins-official |
| skills | Skill creation tools - scaffold, lint, and manage Claude Code skills | /plugin marketplace add anthropics/skills |
Community Tools
| Plugin | Description | Install |
|--------|-------------|---------|
| beads | Dependency-aware issue tracker - issues chained like beads with blocking relationships | curl -fsSL https://raw.githubusercontent.com/steveyegge/beads/main/install.sh \| bash |
| beads_viewer | Terminal UI for Beads with DAG visualization, graph metrics, and AI agent "robot mode" | curl -fsSL https://raw.githubusercontent.com/Dicklesworthstone/beads_viewer/main/install.sh \| bash |
| claude-hud | Real-time session insights - context window usage, active tools, todo progress, git status in statusline | /plugin marketplace add jarrodwatts/claude-hud |
| browser-cdp | Browser automation via Chrome DevTools Protocol - control Chrome, Brave, Edge with real profiles | npm install -g browser-cdp or bunx browser-cdp |
| agent-browser | Fast Rust CLI for AI agents - semantic locators, accessibility snapshots, session isolation | npm install -g agent-browser && agent-browser install |
Recommended Settings
Add Claude Code environment variables to your shell profile:
# Option 1: Use the installer (menu option 4) curl -fsSL https://raw.githubusercontent.com/majesticlabs-dev/majestic-marketplace/master/install.sh | bash # Option 2: Manual - append to ~/.zshrc or ~/.bashrc curl -fsSL https://raw.githubusercontent.com/majesticlabs-dev/majestic-marketplace/master/instructions/shell-settings.sh >> ~/.zshrc source ~/.zshrc
| Setting | Description |
|---------|-------------|
| ENABLE_BACKGROUND_TASKS | Enables background task execution |
| FORCE_AUTO_BACKGROUND_TASKS | Automatically runs eligible tasks in background |
| CLAUDE_BASH_MAINTAIN_PROJECT_WORKING_DIR | Keeps bash commands in project directory |
| cly alias | alias cly="claude --dangerously-skip-permissions" - auto-approve for trusted projects |
MCP Servers
Add the Sequential Thinking MCP server for enhanced reasoning:
claude mcp add sequential-thinking -s local -- npx -y @modelcontextprotocol/server-sequential-thinking
Updating Marketplace & Plugins
Update Everything
/plugin marketplace update majestic-marketplace
This pulls the latest changes from the marketplace repository and updates all installed plugins.
Update Individual Plugin
/plugin update majestic-engineer
Check Installed Versions
/plugin list
Output Styles
Configure how Claude Code communicates with you using the /output-style command:
/output-style

| Style | Description | |-------|-------------| | Bullet Points | Hierarchical bullet points for quick scanning | | Explanatory | Claude explains its implementation choices and codebase patterns | | GenUI | Generative UI with embedded modern styling | | HTML Structured | Clean semantic HTML with proper structure | | Learning | Claude pauses and asks you to write small pieces of code for hands-on practice | | Markdown Focused | Full markdown features for maximum readability | | Table Based | Markdown tables for better organization and scanning | | TTS Summary | Audio task completion announcements with TTS | | Ultra Concise | Minimal words maximum speed direct actions | | YAML Structured | Machine-readable YAML format for structured data and easy parsing |
See Output Styles for full documentation and customization.
Project Setup
Initialize Project Configuration
After installing plugins, run this command in your project to set up AI agent configuration:
/majestic:init
This creates:
- AGENTS.md - Human-readable guidance for Claude (WHAT/WHY/HOW framework)
- .agents.yml - Machine-readable config for commands
- CLAUDE.md - Symlink to AGENTS.md
.agents.yml Config
The .agents.yml file stores project-specific settings that commands read automatically:
# .agents.yml - Project configuration for Claude Code commands default_branch: main tech_stack: rails # rails | python | generic app_status: development # development | production task_management: github # github | linear | beads | file | none workflow: worktrees # worktrees | branches branch_naming: type/issue-desc # feature/desc | issue-desc | type/issue-desc | user/desc lessons_path: .agents-os/lessons/
Commands like /majestic:code-review, /git:create-pr, and /majestic:build-task read this config automatically.
Custom Config Path: Override the config filename with AGENTS_CONFIG environment variable:
export AGENTS_CONFIG=".my-project-config.yml"
CLAUDE.md Instruction Modules
Add pre-built coding guidelines to your CLAUDE.md using the installer:
bash -c "$(curl -fsSL https://raw.githubusercontent.com/majesticlabs-dev/majestic-marketplace/master/install.sh)"
Available modules:
- critical-thinking - Challenge assumptions, prioritize accuracy over validation
- anti-overengineering - YAGNI, KISS, DRY - avoid unnecessary complexity
- analyze-vs-implement - Distinguish suggestion from action
- investigation-first - Understand before proposing solutions
Resources
| Resource | Description | |----------|-------------| | Marketplace Tutorial | Interactive slide deck for getting started | | Output Styles | Pre-built formatting guides for consistent responses | | Instructions | CLAUDE.md instruct
优点
- 促进工程师之间的协作。
- 通过 AI 代理简化编码工作流程。
- 提供多种插件以满足不同需求。
缺点
- 新用户可能需要学习曲线。
- 对外部插件的依赖可能会使设置复杂。
- 对某些编码环境的支持有限。
相关技能
免责声明:本内容来源于 GitHub 开源项目,仅供展示和评分分析使用。
版权归原作者所有 majesticlabs-dev.
