Claude Plugins, Skills, and MCP

One Practical Way to Understand Plugins, Skills, and MCP

Teams often use terms like skills, plugins, and MCP together, but they may refer to different layers of a workflow depending on the platform.

• Skill: a set of instructions, examples, and supporting files that shape how a task is performed.
• MCP server: a tool layer that helps connect a workflow to external systems or data sources.
• Plugin-like package: an installation or packaging model that may combine instructions, tool access, and configuration in one reusable unit.

Important: This is a practical explanatory framework, not a claim of one official packaging standard across every Claude-compatible environment.

• Example: a plugin-style package may bundle an MCP server for access and one or more skills for behavior. Treat the package names here as conceptual examples unless they are linked to an official source.

Common Plugin and Integration Patterns for Developers

Installation Guide

Important: The command patterns and configuration examples below are illustrative only. Verify current CLI syntax, package names, and supported configuration formats in the official documentation for your environment before using them in production.

• You can install plugins via the command line or a visual interface.
• Method 1: CLI-based installation pattern
• Some environments may use command-based installation flows for plugin or integration management.

# Example only — verify actual syntax in official docs
 /plugin search example-plugin
 /plugin install example-plugin
 /plugin list
 

• Method 2: Configuration-based setup
• Some teams prefer to document workflow dependencies in a shared configuration file, but the exact schema depends on the platform.

{
 "plugins": {
 "example-plugin-a": "latest",
 "example-plugin-b": "1.x"
 }
 }

One Possible Structure for an Internal Plugin Package

• Building a plugin allows you to share your custom tools and skills with the world. A plugin is simply a directory with a plugin.json manifest.
• File Structure:

my-plugin/ ├── plugin.json # Manifest (Name, Version, Permissions) ├── skills/ # Folder containing SKILL.md files │ └── my-skill.md └── server/ # (Optional) MCP Server code └── index.py

• plugin.json Example:

{ "name": "my-custom-plugin", "version": "1.0.0", "description": "Internal tools for Acme Corp.", "permissions": ["filesystem:read", "network:internal-api"] }

Visual Plugin Management in Verdent

Managing CLI plugins can get messy. Verdent provides an "App Store" experience for your AI.

• Visual install flow: some environments let you browse available packages and install them without editing config files manually.
• Permission Control: Visually toggle permissions. Want to disable the "Slack Plugin" from posting messages? Just uncheck the "Write" box.
• auto-update: Verdent checks for plugin updates in the background and notifies you of security patches.
• Workspace Isolation: Install "Data Science Plugins" only for your Python projects and "Web Dev Plugins" for your React projects, keeping your environment clean.

Note: The plugin IDs below are illustrative example Plugin IDs. Check the current official documentation or linked repository source for actual package names and availability.

Frequently Asked Questions