Claude Code Plugins: Guide and Security

Claude Code plugins are installable directories that package skills, agents, commands, hooks, MCP servers, language servers, and executables into a single configurable unit, typically distributed through marketplaces and managed by a manifest. While skills and agents primarily provide instructions and workflows, hooks, monitors, and server processes can execute with the user’s privileges and may operate outside Claude Code’s permission rules and sandbox, making plugin installation a software supply-chain trust decision rather than merely a prompt configuration. The overview describes official, community, and third-party marketplaces, notes that the official catalog contained 317 plugins as of November 9, 2026, and explains installation scopes for individual users, projects, and local repositories. It recommends reviewing marketplace sources, manifests, hooks, MCP definitions, executable files, repository practices, and version pins before installation, citing incidents including Plugin4Shell’s commit-pin bypass, a Hookify trust-channel issue, and malicious Claude Code IDE extensions in Cursor’s marketplace. For organizations, it highlights managed settings that can allowlist marketplaces, enable or block plugins, prevent sideloading, limit hooks, and audit plugin activity, while arguing that runtime monitoring may be needed because static policies cannot fully control the actions or changing outputs of approved plugins.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.