Claude Code plugins

A Claude Code plugin is an installable bundle. It can contain skills, subagents, hooks, MCP server configuration, and LSP integrations. A skill is one reusable workflow inside that ecosystem; do not use the terms interchangeably.

Official guides: Discover plugins and Plugin marketplaces.

Browse and install

Open the plugin panel:

/plugin

Add Anthropic’s official marketplace when it is not already configured:

/plugin marketplace add anthropics/claude-plugins-official

Install a plugin:

/plugin install github@claude-plugins-official

The scripted equivalent is:

claude plugin install github@claude-plugins-official

If the install summary asks you to reload newly installed components, run:

/reload-plugins

Already-active plugins do not need a reload after every use.

Inspect before trusting

A plugin can execute hooks, add tools, and connect external systems. Before enabling one:

• inspect its source and declared components,
• check which MCP servers and credentials it expects,
• review hook scripts as executable code,
• pin or govern marketplaces for shared environments,
• test it in a disposable repository before organization-wide rollout.

Organization controls

Managed settings can constrain plugin sources and marketplaces with controls such as strictKnownMarketplaces, blockedMarketplaces, disableSideloadFlags, and disableCommandPluginSources. Platform teams should prefer reviewed internal or allowlisted marketplaces over asking every developer to make a fresh trust decision.

When to use which mechanism