omni-agents

Translations: Português | Español

A centralized, tool-agnostic hub to version, manage, and distribute specialized subagents, development rules, and modular skill libraries across multiple AI coding assistants.

Supported Tools

omni-agents unifies context management and compiles directly into the native format expected by each assistant:

Repository Structure

omni-agents/
├── agents/                          # Specialized subagent personas
│   ├── accessibility-reviewer.md
│   ├── code-reviewer.md
│   └── security-auditor.md
├── rules/                           # Modular rule profiles (one AGENTS.md per folder)
│   ├── general/                     # General engineering guidelines (English)
│   │   └── AGENTS.md
│   └── pt-br-dev/                   # Brazilian Portuguese development guidelines
│       └── AGENTS.md
├── scripts/                         # Automation configurator and target adapters
│   ├── configure_workspace.py       # Interactive TUI and CLI orchestrator
│   ├── i18n.py                      # Decoupled internationalization engine
│   ├── languages/                   # Localized translation catalogs (en.json, pt.json...)
│   └── targets/                     # Specialized target adapter modules
│       ├── __init__.py              # Central target registry
│       ├── base.py                  # Adapter contracts and filesystem utilities
│       ├── antigravity.py           # Google Antigravity adapter
│       ├── claude.py                # Claude Code adapter (CLAUDE.md)
│       ├── cursor.py                # Cursor IDE adapter (.cursor/rules/*.mdc)
│       ├── copilot.py               # GitHub Copilot adapter
│       └── universal.py             # Universal, Kiro, OpenCode, and Codex adapter
├── skills/
│   ├── global/                      # GLOBAL CORE (~800 tokens) - Essential for all projects
│   │   ├── coding-standards/        # Technical excellence, Clean Code, and engineering patterns
│   │   ├── comunicacao-clara/       # Clear, unpacked communication and concise synthesis
│   │   ├── grug-brained-dev/        # Anti-overengineering and simplicity mindset
│   │   ├── reducing-entropy/        # Combating technical debt and deleting dead code
│   │   └── researching-codebases/   # Multi-agent codebase research and navigation
│   ├── planning/                    # LIBRARY: Planning, auditing, and roadmaps
│   ├── docs/                        # LIBRARY: Technical writing and documentation
│   ├── frontend/                    # LIBRARY: UI design, visual standards, and Tailwind CSS
│   ├── meta/                        # LIBRARY: Prompt engineering and customization authoring
│   └── stacks/                      # LIBRARY: Frameworks, languages, and infrastructure
├── docs/
│   └── translations/                # Translated documentation
│       ├── README.pt-BR.md
│       └── README.es.md
└── README.md                        # Primary documentation in English

Contents

1. Subagents (agents/)

• accessibility-reviewer.md: Digital accessibility (a11y) specialist covering WCAG 2.1/2.2 (A, AA, AAA) and WAI-ARIA guidelines.
• code-reviewer.md: Static code analysis, Clean Code, readability, and solid architecture reviewer.
• security-auditor.md: Application security specialist covering OWASP Top 10 and secure coding practices.

2. Rule Profiles (rules/)

Organized into dedicated profile folders, each containing its own AGENTS.md contract:

• pt-br-dev/AGENTS.md: Engineering guidelines, security standards, Brazilian Portuguese communication rules, strict accentuation, emoji prohibition, and Tailwind CSS.
• general/AGENTS.md: Universal guidelines for international projects in English, clean architecture, security, and modern frontend practices.

3. Modular Skill Library (skills/)

To optimize token consumption and prevent model choice hesitation:

1. Global Core (skills/global/): 5 essential skills establishing foundational pair-programming behavior across any project (~800 tokens total).
2. On-Demand Library (skills/{planning,docs,frontend,meta,stacks}/): Specialized skills selectively mounted only in relevant repositories.

Installation

Install omni-agents as a global CLI tool on your system with a single command:

One-Line Installers

• Windows (PowerShell):

  irm https://raw.githubusercontent.com/TheCheepeer/omni-agents/main/install.ps1 | iex
  

• Linux / macOS:

  curl -fsSL https://raw.githubusercontent.com/TheCheepeer/omni-agents/main/install.sh | bash
  

Via Package Managers (pipx / uv)

pipx install omni-agents-cli
# Or using uv:
uv tool install omni-agents-cli

Once installed, execute omni-agents (or simply omni) anywhere in your terminal.

Cross-Platform Configuration & Automation

The tool can be executed globally via the omni-agents command (or omni), or directly through the source script scripts/configure_workspace.py when working in the cloned repository (zero external dependencies).

1. Interactive Mode (TUI / GUI)

# Launch interactive menu globally:
omni-agents

# Or within the cloned repository:
python scripts/configure_workspace.py

# Or specify target project directory directly:
omni-agents /path/to/project
omni-agents .

Interactive Menu Layout:

=================================================================
  MULTI-TOOL AGENT, RULES & SKILLS CONFIGURATOR
=================================================================
  Target Workspace:   [/path/to/project]
  Active Tools:       antigravity, cursor, claude
-----------------------------------------------------------------
  [t] Select Target Tools (Antigravity, Cursor, Claude...)
  [1] Global Machine Configuration (Antigravity & Claude)
  [2] Subagents for Workspace
  [3] Rules for Workspace
  [4] Modular Skills for Workspace
  [s] Synchronize All (Sync active tools in batch)
  [c] Clean / Uninstall by Tool
  [l] Language / Idioma: [EN | PT-BR]
  [5] Exit
-----------------------------------------------------------------
  [w] Set / Change Target Workspace
=================================================================

2. Command-Line Interface (CLI / Automation)

Execute commands directly within scripts or CI/CD pipelines:

# List all supported tools:
python scripts/configure_workspace.py --list-tools

# Synchronize workspace across all active tools:
python scripts/configure_workspace.py /path/to/project --sync

# Configure for a specific tool:
python scripts/configure_workspace.py /path/to/project --tool cursor --sync
python scripts/configure_workspace.py /path/to/project --tool claude --sync
python scripts/configure_workspace.py /path/to/project --tool copilot --sync

# Configure for all tools simultaneously (Multi-Tool):
python scripts/configure_workspace.py /path/to/project --tool all --sync

# Remove configuration for a single tool without affecting others:
python scripts/configure_workspace.py /path/to/project --tool cursor --clean
python scripts/configure_workspace.py /path/to/project --clean  # Remove all

# Apply global machine configuration:
python scripts/configure_workspace.py --global
python scripts/configure_workspace.py --tool claude --global

# Force UI language:
python scripts/configure_workspace.py --lang en
python scripts/configure_workspace.py --lang pt

3. Strategic Scenarios

Multi-Tool Team Workflow

In teams where different developers use different environments (e.g., developer A uses Cursor, developer B uses VS Code with Copilot, and developer C uses Antigravity), use --tool all. The script exports native configuration files for each assistant from a single source of truth, guaranteeing project-wide consistency.

Fast Synchronization (Sync)

When updating rules or skills in this repository, run --sync on any target project to instantly regenerate consolidated files (CLAUDE.md, copilot-instructions.md, .cursor/rules/, etc.).

State Persistence

Workspace preferences are saved in .agents/workspace_state.json (automatically added to .gitignore), allowing you to inspect and modify active targets at any time.

Documentation

For in-depth technical documentation and developer guides, consult:

• Architecture & Design: Layered component resolution, filesystem linking engine, and target adapter patterns.
• CLI Reference: Complete command-line syntax, flags, interactive TUI navigation, and headless/CI automation.
• Configuration & Extensions: Schema for config.json, personal overrides in custom/, remote catalog syncing, and updater.
• Contributing Standards: Local environment setup, engineering guidelines, adding new target adapters, and localization.

Credits and Acknowledgements

Several skills included in this repository were adapted from:

• agent-skills by Josh Thomas (MIT License).
• The Grug Brained Developer by Colin McDonnell.
• Diátaxis Documentation Framework by Daniele Procida.

See the NOTICE file for formal third-party license and copyright declarations.

License

Distributed under the Apache 2.0 License. See LICENSE for details.