Gateway Mcp

CI codecov Quality Gate Status Node TypeScript License: MIT

A toolkit for testing and packaging portable AI agents that work across various LLMs, frameworks, deployment targets, and orchestrators.

Features

Agent Development

• 🤖 Multi-LLM Support - Build agents that work with Claude, GPT, and other LLMs
• 🔌 Framework Agnostic - Support for various agent frameworks
• 📦 Portable Packaging - Deploy across different orchestrators and platforms
• ✅ Agent Testing - Comprehensive testing for agent behaviors and interactions
• 🔍 Agent Skills Validation - Audit skills for quality, compatibility, and best practices
• 🩺 Environment Diagnostics - Doctor command checks setup and health

Skills Distribution

• 📦 VAT Distribution Standard - Package-based distribution for skills, agents, and tools
• 🔨 Build Infrastructure - vat skills build creates distributable skill packages
• ⬇️ Smart Installation - Install skills from npm, local directories, or zip files
• 📋 Registry Tracking - Track installed skills for updates and management
• 🎯 Multi-Artifact Support - Distribute skills, agents, pure functions, and runtimes together

Install skills:

vat skills install npm:@vibe-agent-toolkit/vat-example-cat-agents
vat skills install ./path/to/local/package
vat skills list --installed

See Distributing VAT Skills Guide for creating your own distributable packages.

Claude Plugin Marketplace

VAT skills are available as a Claude Code plugin — install them directly from GitHub, no npm needed:

claude plugin marketplace add jdutton/vibe-agent-toolkit#claude-marketplace
claude plugin install vibe-agent-toolkit@vat-skills

Or from within a Claude Code session:

/plugin marketplace add jdutton/vibe-agent-toolkit#claude-marketplace
/plugin install vibe-agent-toolkit@vat-skills

For project-scope (shared with team via .claude/settings.json), add --scope project:

claude plugin marketplace add jdutton/vibe-agent-toolkit#claude-marketplace --scope project

This gives Claude Code 8 skills for building, packaging, and distributing agent skills. See the marketplace branch for the full skill list.

Pre-release channel: To track the latest pre-release builds, use #claude-marketplace-next instead.

Build your own marketplace: VAT can publish your skills as a Claude plugin marketplace too. See the Marketplace Distribution Guide for how marketplace publishing works — from project structure through build, validation, publishing, and testing.

Plugin & Marketplace Audit

Comprehensive validation for Claude plugins, marketplaces, and configurations:

# Audit user-level Claude plugins
vat audit --user

# Audit specific plugin
vat audit ./my-plugin

# Recursive scan for all resources
vat audit ./resources --recursive

Features:

• Auto-detects resource type (plugin, marketplace, skill, registry)
• Schema validation using Zod
• Cache staleness detection using checksums
• Hierarchical output showing relationships
• Cross-platform support

See Audit Documentation for complete reference.

Development Infrastructure

• 🚀 Bun - Fast package manager and runtime
• 📦 Monorepo - Workspace-based package management for agent components
• 🔒 TypeScript - Strict type checking with composite projects
• ✅ Vitest - Fast unit, integration, and system testing
• 🔍 ESLint - Maximum strictness with sonarjs, unicorn, and security plugins
• 📊 Code Coverage - 80% minimum threshold with Codecov integration
• 🎯 vibe-validate - Git-aware validation orchestration
• 🔧 Cross-Platform - Tested on Windows, macOS, and Linux
• 🤖 CI/CD - GitHub Actions with Node 22/24 matrix testing

Quick Start

CLI Installation

Install the CLI globally:

npm install -g vibe-agent-toolkit

Scan and validate markdown resources:

vat resources scan docs/
vat resources validate docs/

Audit and import Agent Skills:

vat agent audit my-skill/SKILL.md
vat agent import my-skill/SKILL.md

Diagnose project setup:

vat doctor

See CLI Reference for complete documentation (or run vat --help --verbose).

Development Setup

Prerequisites:

• Bun (latest version)
• Node.js 22 or 24
• Git

Installation:

# Clone the repository
git clone https://github.com/jdutton/vibe-agent-toolkit.git
cd vibe-agent-toolkit

# Install dependencies
bun install

# Build all packages
bun run build

Development

# Run full validation (recommended)
bunx vv validate

# Or run test suites individually:
bun run test:unit          # Unit tests only
bun run test:watch         # Watch mode for development
bun run test:integration   # Integration tests
bun run test:system        # System/e2e tests

# Lint code
bun run lint

# Type check
bun run typecheck

# Validate all markdown links (dogfooding)
bun run validate-links

🔗 Dogfooding: Link Validation

This project uses its own @vibe-agent-toolkit/resources package to validate all markdown links in the repository. This ensures our documentation stays accurate and all links remain valid.

Run link validation:

bun run validate-links

What it checks:

• ✅ All local file links (relative paths to other markdown files)
• ✅ All heading anchor links (links to sections within files)
• ℹ️ External URLs are noted but not validated
• 🚫 Test fixtures with intentionally broken links are excluded

Runs automatically in CI/CD as part of the System Tests phase, ensuring broken links are caught before merge.

External URL Validation

Optionally validate external URLs with caching:

vat resources validate docs/ --check-external-urls

See External URL Validation for details.

Project Structure

vibe-agent-toolkit/
├── packages/              # Published packages
│   └── example-utils/    # Example package (replace with your packages)
│       ├── src/          # Source code
│       ├── test/         # Tests
│       ├── package.json
│       └── tsconfig.json
├── tools/                # Development tools (TypeScript)
│   ├── common.ts         # Shared utilities
│   ├── duplication-check.ts
│   ├── jscpd-check-new.ts
│   └── jscpd-update-baseline.ts
├── docs/                 # Documentation
├── .github/              # CI/CD workflows
│   └── workflows/
│       └── validate.yml  # Validation pipeline
├── eslint.config.js      # ESLint configuration (strict!)
├── vitest.config.ts      # Unit test configuration
├── vitest.integration.config.ts  # Integration test config
├── vitest.system.config.ts       # System test config
├── tsconfig.json         # Root TypeScript config
├── tsconfig.base.json    # Base config for packages
├── vibe-validate.config.yaml  # Validation orchestration
├── package.json          # Root package.json
├── CLAUDE.md            # AI assistant guidelines
└── README.md            # This file

Adding a New Package

1. Create the package directory:

mkdir -p packages/my-package/src packages/my-package/test

2. Create packages/my-package/package.json:

{
  "name": "@your-org/my-package",
  "version": "0.1.0",
  "type": "module",
  "main": "./dist/index.js",
  "types": "./dist/index.d.ts",
  "exports": {
    ".": {
      "types": "./dist/index.d.ts",
      "import": "./dist/index.js"
    }
  },
  "scripts": {
    "build": "tsc",
    "test": "vitest run",
    "test:watch": "vitest",
    "typecheck": "tsc --noEmit",
    "clean": "rm -rf dist *.tsbuildinfo"
  },
  "devDependencies": {
    "typescript": "^5.9.3",
    "vitest": "^3.2.4"
  }
}

3. Create packages/my-package/tsconfig.json:

{
  "extends": "../../tsconfig.base.json",
  "compilerOptions": {
    "outDir": "./dist",
    "rootDir": "./src"
  },
  "include": ["src/**/*.ts"],
  "exclude": ["node_modules", "dist", "**/*.test.ts"]
}

4. Add to root tsconfig.json references:

{
  "references": [
    { "path": "./packages/example-utils" },
    { "path": "./packages/my-package" }  // Add this
  ]
}

5. Create source and tests:

# Create your source files
touch packages/my-package/src/index.ts

# Create tests
touch packages/my-package/test/index.test.ts

6. Install and build:

bun install
bun run build

Testing Strategy

Test Types

The toolkit supports three levels of testing:

Unit Tests (*.test.ts)

• Test individual functions/classes
• Fast execution (< 100ms)
• Mock external dependencies
• Run with: bun test

Integration Tests (*.integration.test.ts)

• Test multiple modules together
• May use real dependencies
• Medium execution (< 5s)
• Run with: bun test:integration

System Tests (*.system.test.ts)

• End-to-end workflows
• Real external services
• Longer execution (< 30s)
• Run with: bun test:system

Test File Organization

packages/my-package/
├── src/
│   ├── utils.ts
│   └── utils.test.ts              # Unit tests (co-located)
└── test/
    ├── api.test.ts                # Unit tests
    ├── integration/
    │   └── workflow.integration.test.ts
    └── system/
        └── e2e.system.test.ts

Quality Standards

ESLint

The toolkit enforces strict linting rules:

• Zero warnings policy: --max-warnings=0
• SonarJS: Catches code smells and bugs
• Unicorn: Modern JavaScript patterns
• Security: Prevents common vulnerabilities
• Import organization: Alphabetical with groups

All code (src, tests, tools) is held to the same standards.

Code Coverage

Minimum thresholds enforced:

• Statements: 80%
• Branches: 80%
• Functions: 80%
• Lines: 80%

Code Duplication

Uses jscpd with a baseline approach:

• Existing duplication is baselined
• New duplication blocks commits
• Update baseline: bun run duplication-update-baseline

TypeScript

Strict mode enabled with additional checks:

• noUncheckedIndexedAccess
• noImplicitOverride
• exactOptionalPropertyTypes

Validation Pipeline

The toolkit uses vibe-validate for git-aware validation:

# Run all validation checks
bun run validate

# Pre-commit check (fast)
bun run pre-commit

Validation runs:

1. TypeScript type checking
2. ESLint (strict)
3. Code duplication check
4. Build
5. Unit tests
6. Integration tests (optional)
7. Coverage report
8. System tests (optional)

CI/CD

GitHub Actions workflows run on every push/PR:

Validation Pipeline

• Matrix Testing: Node 22/24 × Ubuntu/Windows
• Full Validation: All checks must pass
• Parallel Execution: Faster feedback
• See .github/workflows/validate.yml

Coverage Reporting

• Codecov Integration: Automatic coverage upload
• 80% Threshold: Project and patch coverage targets
• PR Comments: Coverage diff on pull requests
• See .github/workflows/coverage.yml

Setting up Codecov

1. Sign up at codecov.io with your GitHub account
2. Enable your repository
3. Add CODECOV_TOKEN to GitHub repository secrets
4. Coverage reports will be uploaded automatically on push/PR
5. Update badge URLs in README.md with your repository details

Pre-commit Hooks

Husky is configured to run validation before commits:

# Install hooks
bun install  # Runs automatically via "prepare" script

# Manually run pre-commit checks
bun run pre-commit

Publishing Packages

When you're ready to publish to npm:

1. Update package versions
2. Build packages: bun run build
3. Test thoroughly: bun run validate
4. Publish: npm publish (or create publish scripts)

Customization

Changing the Organization Scope

Replace @vibe-agent-toolkit with your org name:

# Update package names in:
# - packages/*/package.json
# - Import statements

Adjusting Quality Thresholds

Edit these files:

• vitest.config.ts - Coverage thresholds
• eslint.config.js - Linting rules
• vibe-validate.config.yaml - Validation steps

Adding Tools

Add scripts to packages/dev-tools/src/ directory:

• Use TypeScript (cross-platform)
• Import from common.ts
• Add to package.json scripts

Validation

The project includes comprehensive validation checks:

# Run all validation checks
bun run validate

# Run pre-commit checks
bun run pre-commit

The example package (packages/example-utils/) demonstrates testing patterns and will be replaced with agent-specific packages.

Learn More

• Getting Started - Detailed setup guide
• Documentation - Full documentation index
• Bun Documentation
• Vitest Documentation
• vibe-validate
• TypeScript

License

MIT - see LICENSE file

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Run bun run validate to ensure quality
5. Submit a pull request

---

Built with ❤️ using Bun, TypeScript, and modern tooling