🧠 LMAgent: The Universal AI Agent Runtime

██╗     ███╗   ███╗ █████╗  ██████╗ ███████╗███╗   ██╗████████╗
██║     ████╗ ████║██╔══██╗██╔════╝ ██╔════╝████╗  ██║╚══██╔══╝
██║     ██╔████╔██║███████║██║  ███╗█████╗  ██╔██╗ ██║   ██║
██║     ██║╚██╔╝██║██╔══██║██║   ██║██╔══╝  ██║╚██╗██║   ██║
███████╗██║ ╚═╝ ██║██║  ██║╚██████╔╝███████╗██║ ╚████║   ██║
╚══════╝╚═╝     ╚═╝╚═╝  ╚═╝ ╚═════╝ ╚══════╝╚═╝  ╚═══╝   ╚═╝
                                                  by QuBit

"Separate the reasoning from the execution." LMAgent is the foundational runtime that empowers your AI Agents with standardized Skills, Rules, and Workflows across 37 supported agents (Cursor, Claude Code, Windsurf, Gemini CLI, Cline, Roo, Copilot, and many more).

⚡ One Command. Any Agent.

npx @qubiit/lmagent@latest

That's it. No global install required. LMAgent will:

1. Auto-detect which AI agents you have installed on your system
2. Pre-select the detected agents for installation
3. Deploy skills, rules, and workflows strictly to your project directory (Universal Brain: .agents/)
4. Generate the entry point file so each agent auto-invokes the framework

Requires Node.js ≥ 22. Works on macOS, Linux, and Windows. V3.4.0 features a unified install command and centralized .agents/ architecture.

🚀 Full Setup Guide

One Command — Install Everything

npx @qubiit/lmagent@latest

This single command:

1. Deploys AGENTS.md to your project root (the universal entry point)
2. Auto-detects which AI agents you have in the project
3. Installs all skills, rules, workflows, memory, config, and docs to .agents/
4. Generates lightweight bridge files for each detected agent
5. Deploys agent-specific config files (CLAUDE.md, GEMINI.md) only when needed

lmagent init, lmagent install, and lmagent update all run the same unified flow.

Verify

npx @qubiit/lmagent@latest doctor

🏛️ Architecture

LMAgent uses a Hub & Spoke model: one universal brain (.agents/) that feeds all agents.

Your Project/
├── .agents/                    # ← UNIVERSAL BRAIN (source of truth)
│   ├── rules/                  # 11 behavioral rules
│   ├── skills/                 # 31 specialized roles
│   ├── workflows/              # 13 SOPs (Standard Operating Procedures)
│   ├── memory/                 # 5 persistent context files
│   ├── templates/              # Project scaffolds & agent config templates
│   ├── config/                 # Framework settings
│   └── docs/                   # Extended documentation
│
├── AGENTS.md                   # ← UNIVERSAL ENTRY POINT (read by all agents)
├── CLAUDE.md                   # ← Only if Claude Code is detected
├── GEMINI.md                   # ← Only if Gemini CLI / Antigravity is detected
│
├── .cursor/rules/00-lmagent.mdc  # ← Lightweight bridge (points to AGENTS.md)
├── .windsurf/rules/lmagent.md    # ← Lightweight bridge (points to AGENTS.md)
└── ...                           # (one bridge file per installed agent)

How auto-invocation works

Each agent reads a specific file when it starts. LMAgent generates that file automatically:

| Agent | Entry Point Generated | |:---|:---| | Cursor | AGENTS.md (native) + .cursor/rules/00-lmagent.mdc | | Claude Code | CLAUDE.md | | Gemini CLI / Antigravity | GEMINI.md | | Windsurf | .windsurf/rules/lmagent.md | | Cline | .clinerules/00-lmagent.md | | Roo Code | AGENTS.md (native) + .roo/rules/00-lmagent.md | | VSCode Copilot | .github/copilot-instructions.md | | Zed | AGENTS.md (native) + .rules/lmagent.md | | Goose | .goosehints | | Continue | .continue/continuerules | | Junie | .junie/guidelines.md | | OpenHands | .openhands/microagents/repo.md | | Codex CLI | AGENTS.md (native) | | All others | 00-lmagent.md in their rulesDir |

All entry points point to AGENTS.md — the single source of truth.

🧩 Skills Catalog (31 Skills)

Activate any skill by typing its trigger in the chat:

🎯 Management & Architecture

| Trigger | Skill | Description | |:---|:---|:---| | /orch | orchestrator | High-level task planning and agent coordination | | /pm | product-manager | PRD generation, user stories, roadmap planning | | /arch | architect | System design, scalable patterns, cloud infrastructure | | /lead | tech-lead | Code review, technical decisions, mentoring | | /sm | scrum-master | Agile ceremonies, sprint planning, retrospectives | | /doc | technical-writer | Documentation, API refs, user guides |

🔧 Engineering

| Trigger | Skill | Description | |:---|:---|:---| | /dev | backend-engineer | APIs, database schema, authentication | | /front | frontend-engineer | React, Next.js, Tailwind, state management | | /mobile | mobile-engineer | React Native, Expo, iOS/Android pipelines | | /data | data-engineer | ETL pipelines, SQL optimization, data warehousing | | /devops | devops-engineer | CI/CD, Docker, Kubernetes, Terraform | | /sec | security-analyst | Vulnerability scanning, OWASP, code auditing | | /test | qa-engineer | E2E testing (Playwright), unit tests, QA strategy | | /review | code-reviewer | Static analysis, logic verification, clean code | | /api | api-designer | REST/GraphQL contracts, OpenAPI standards | | /supa | supabase-expert | Supabase architecture, RLS, Edge Functions | | /git | git-workflow | Branch management, conventional commits, releases | | /web | browser-agent | Web automation, scraping, UI testing | | /seo | seo-auditor | Technical SEO, Core Web Vitals, accessibility |

⚡ Specialized & AI

| Trigger | Skill | Description | |:---|:---|:---| | /fix | systematic-debugger | Methodical 4-phase debugging (RCA) | | /perf | performance-engineer | Profiling, load testing, optimization | | /ux | ux-ui-designer | Design systems, user flow, prototyping | | /agent | ai-agent-engineer | Building LLM-based systems & MCP servers | | /auto | automation-engineer | n8n workflows, Zapier, scripting | | /prompt | prompt-engineer | Optimizing system prompts and LLM interactions | | /mcp | mcp-builder | Build MCP servers, tools, and resources | | /pdf | document-generator | Programmatic PDF/DOCX/XLSX generation |

🧠 Methodologies

| Trigger | Skill | Description | |:---|:---|:---| | /bmad | bmad-methodology | Scale-Adaptive Intelligence, complexity classification | | /sdd | spec-driven-dev | Spec-first development pipeline | | /swe | swe-agent | Autonomous issue resolution, trajectory logging | | /test-s | testing-strategist | Test planning, pyramid strategy, coverage goals |

🌍 37 Supported Agents

| Agent | Config Path | Entry Point | |:---|:---|:---| | Cursor | .cursor/ | AGENTS.md (native) + .cursor/rules/00-lmagent.mdc | | Windsurf | .windsurf/ | .windsurf/rules/lmagent.md | | Cline | .clinerules/ | .clinerules/00-lmagent.md | | Roo Code | .roo/ | .roo/rules/00-lmagent.md | | VSCode Copilot | .github/ | .github/copilot-instructions.md | | Trae | .trae/ | .trae/rules/lmagent.md | | Trae CN | .trae-cn/ | .trae-cn/rules/lmagent.md | | Claude Code | .claude/ | CLAUDE.md | | Zed | .rules/ | AGENTS.md (native) + .rules/lmagent.md | | Amp / Kimi / Replit | .agents/ | AGENTS.md (native) | | Antigravity | .agent/ | GEMINI.md | | Augment | .augment/ | .augment/rules/00-lmagent.md | | Gemini CLI | .gemini/ | GEMINI.md | | OpenClaw / Envoid | rules/ | openclaw.json | | CodeBuddy | .codebuddy/ | .codebuddy/rules/00-lmagent.md | | Codex CLI | .codex/ | AGENTS.md | | Command Code | .commandcode/ | .commandcode/rules/00-lmagent.md | | Continue | .continue/ | .continue/continuerules | | Crush | .crush/ | .crush/rules/00-lmagent.md | | Droid | .factory/ | .factory/rules/00-lmagent.md | | Goose | .goose/ | .goosehints | | Junie | .junie/ | .junie/guidelines.md | | iFlow CLI | .iflow/ | .iflow/rules/00-lmagent.md | | Kilo Code | .kilocode/ | .kilocode/rules/00-lmagent.md | | Kiro CLI | .kiro/ | .kiro/rules/00-lmagent.md | | Kode | .kode/ | .kode/rules/00-lmagent.md | | MCPJam | .mcpjam/ | .mcpjam/rules/00-lmagent.md | | Mistral Vibe | .vibe/ | .vibe/rules/00-lmagent.md | | Mux | .mux/ | .mux/rules/00-lmagent.md | | OpenCode | .opencode/ | .opencode/rules/00-lmagent.md | | OpenHands | .openhands/ | .openhands/microagents/repo.md | | Pi | .pi/ | .pi/rules/00-lmagent.md | | Qoder | .qoder/ | .qoder/rules/00-lmagent.md | | Qwen Code | .qwen/ | .qwen/rules/00-lmagent.md | | Zencoder | .zencoder/ | .zencoder/rules/00-lmagent.md | | Neovate | .neovate/ | .neovate/rules/00-lmagent.md | | Pochi | .pochi/ | .pochi/rules/00-lmagent.md | | AdaL | .adal/ | .adal/rules/00-lmagent.md |

🛠️ CLI Reference

# Core (all three run the same unified flow)
npx @qubiit/lmagent@latest              # Interactive install (auto-detects agents)
npx @qubiit/lmagent@latest install      # Same as above
npx @qubiit/lmagent@latest init         # Alias for install
npx @qubiit/lmagent@latest update       # Alias for install
npx @qubiit/lmagent@latest uninstall    # Remove all LMAgent files from project
npx @qubiit/lmagent@latest uninstall --all  # Also remove root entry points (CLAUDE.md, etc.)

# Diagnostics
npx @qubiit/lmagent@latest doctor       # Verify project configuration
npx @qubiit/lmagent@latest validate     # Validate integrity of all skills
npx @qubiit/lmagent@latest tokens       # Analyze framework token consumption

# Skills Management
npx @qubiit/lmagent@latest create-skill             # Create a new skill interactively
npx @qubiit/lmagent@latest skills add owner/repo    # Install external skill from GitHub

🛠️ Creating Custom Skills

# Create a new skill interactively
npx @qubiit/lmagent@latest create-skill

# Install an external skill from GitHub
npx @qubiit/lmagent@latest skills add owner/repo-name

# Validate all skills
npx @qubiit/lmagent@latest validate

Skills follow the standard structure:

.agents/skills/my-skill/
├── SKILL.md          # Main instructions (required)
└── ...               # Optional additional files

👨‍💻 Developer Setup

If you are contributing to the framework:

git clone https://github.com/QuBiit0/lmagent.git
cd lmagent
npm install
npm link
lmagent doctor

🤝 Contributing

We welcome contributions to expand the Universal Agent Brain. Check out CONTRIBUTING.md to add new Skills, Rules, or IDE support.