What are Claude Code skills and plugins, and which are worth installing?

A skill is a SKILL.md file of instructions that Claude loads when relevant or that you run as /skill-name. A plugin is a directory that packages skills, subagents, hooks and MCP servers so they install and update as one unit, usually from a marketplace. Worth installing for most developers, as of October 26, 2026: the language-server plugin for your stack (typescript-lsp, pyright-lsp and others), commit-commands, code-review or pr-review-toolkit, security-guidance, and skill-creator to write your own skills — all from Anthropic's official marketplace.

How do I install a Claude Code plugin?

In a Claude Code session run /plugin install @claude-plugins-official, review what it adds and its context cost, then choose a scope: user (every project on your machine), project (everyone working in the repository) or local (you, in this repository only). From the shell, claude plugin install @claude-plugins-official does the same, and claude plugin list shows what is installed. Plugins from other marketplaces need that marketplace added first.

Do plugins slow Claude Code down or cost tokens?

They cost context. For every skill, agent and command Claude can invoke on its own, the name and description sit in context on every turn, even when unused, and the plugin's MCP servers run alongside the session. Plugins in Anthropic's official marketplace show a Context cost estimate before you install; disable plugins you are not using with /plugin or claude plugin disable.

Where do I put my own Claude Code skills?

Personal skills go in ~/.claude/skills/ /SKILL.md and load in every project on your machine. Project skills go in .claude/skills/ /SKILL.md in the repository; commit them so your team gets them. Old .claude/commands/ files keep working — custom commands have been merged into skills.

Quick Answer

How to Install Claude Code Plugins and Skills (2026 Guide)

Published: October 26, 2026

The short answer

Skills are instructions; plugins are packages. A skill is a SKILL.md file that Claude Code loads when it is relevant, or that you run as /skill-name. A plugin bundles several skills, subagents, hooks and MCP servers so they install, update and switch off together. To install one from Anthropic’s official marketplace, type this in a Claude Code session:

/plugin install commit-commands@claude-plugins-official

Review what it adds, pick a scope, and its skills appear as /commit-commands:<skill>. The rest of this guide covers scopes, writing your own skills, the context cost, and which plugins are worth it. Instructions match Anthropic’s Claude Code documentation as of October 26, 2026.

Step 1: Decide whether you need a plugin or a skill

Anthropic’s docs make the distinction simple: skills, subagents, hooks and MCP servers all work on their own. Use a single skill when you keep pasting the same checklist or procedure into chat. Use a plugin when you want someone else’s whole setup in one command with updates, or want to give your own setup to a team.

Claude Code skills follow the Agent Skills open standard, which Anthropic donated to the Agentic AI Foundation on October 22, 2026, so a skill you write can also work in other agents that support the format.

Step 2: Install a plugin and choose a scope

Run /plugin install <name>@<marketplace> — it opens the plugin’s details instead of installing immediately. Run /plugin alone to browse the Discover tab.
Read Will install (commands, agents, skills, hooks, MCP and LSP servers) and, for official plugins, Context cost: what it adds to every turn and what it adds when invoked.
Choose a scope:
- User — every project on this machine.
- Project — written to the repository, so everyone who works there gets it.
- Local — you, in this repository only.
Read the summary. “Plugin is now active” means done; “Run /reload-plugins to apply” means a reload is needed; a bundled MCP server may need Configure on the Installed tab.
Confirm with / (look for /<plugin>:<skill>) or claude plugin list in your shell.

From a script or a marketplace page’s copy button, the shell form is claude plugin install <name>@claude-plugins-official. In VS Code, type /plugins in the Claude Code panel. Cloud sessions at claude.ai/code do not load the plugins installed on your machine.

Step 3: Add your own skills

Create a folder and a SKILL.md in the right place:

Location	Path	Who gets it
Personal	`~/.claude/skills/<name>/SKILL.md`	You, in every project on this machine
Project	`.claude/skills/<name>/SKILL.md`	Everyone in the repository (commit it)
Nested	`<subdir>/.claude/skills/<name>/SKILL.md`	Sessions working in that subdirectory
Plugin	`<plugin>/skills/<name>/SKILL.md`	Wherever the plugin is enabled, as `/plugin:name`

A skill’s body loads only when it is used, so long reference material costs almost nothing until needed — which is why Anthropic suggests moving procedures out of an oversized CLAUDE.md into skills. The official skill-creator plugin will draft and test skills for you.

Step 4: Keep the context cost under control

Every enabled plugin is part of every session: the names and descriptions of its auto-invocable skills and agents sit in context on each turn, its MCP servers run, its hooks fire, and everything it runs, runs as you. So:

Install at project or local scope when a plugin only matters in one repository.
Check the Installed tab, which groups the plugins you have stopped using, and disable them (claude plugin disable).
Run /doctor, which compares unused skills, MCP servers and plugins against their context cost.
Read a third-party plugin’s hooks and MCP servers before installing. Anthropic’s marketplace README warns that it cannot verify what third-party plugins contain. See how to secure coding-agent plugins and skills.

Which plugins are worth installing

All from claude-plugins-official, maintained by Anthropic:

Plugin	What it does	Install if
typescript-lsp / pyright-lsp / gopls-lsp / rust-analyzer-lsp	Language-server code intelligence	Always — pick your language
commit-commands	Commit, push and open PRs	You use git daily
code-review / pr-review-toolkit	Multi-agent PR review with confidence scoring / specialised review agents	You review PRs
security-guidance	Warnings on edits, diff review on stop, commit reviewer	Agent writes production code
skill-creator	Create, improve and measure skills	You want your own skills
claude-md-management	Audit and update CLAUDE.md files	Your CLAUDE.md is growing
session-report	HTML report of token use, cache efficiency and expensive prompts	You watch usage limits

For community plugins, see our best Claude Code plugins ranking, and for how plugins relate to MCP servers, agent plugins vs MCP vs agent skills.

Last verified: October 26, 2026.