Top Claude Code Plugins And Skills In 2027

Claude Code is more useful with a few good plugins and much worse with too many. A plugin bundles skills, slash commands, hooks, subagents and MCP servers into one package that installs with a single command, and Anthropic’s official marketplace, claude-plugins-official, is added to Claude Code automatically. By the autumn of 2026 that marketplace listed roughly 300 plugins, from Anthropic’s own workflow tools to partner integrations and community methodologies.

This list picks the ten Claude Code plugins and skills that close a real gap for working developers: better reviews, safer code, current documentation, working browsers and repeatable workflows. Every entry is in the official marketplace, so each one installs with /plugin install <name>@claude-plugins-official. If you are still choosing an agent, start with our comparison of Claude Code vs Cursor.

Claude Code Plugins At A Glance

Plugin	Publisher	Best For	What It Adds
Superpowers	Jesse Vincent	Multi-step features	Brainstorming, planning, TDD and debugging skills
code-review	Anthropic	Teams reviewing many PRs	Parallel review agents with confidence scoring
security-guidance	Anthropic	Code that touches user input	Warnings on risky edits and a security diff review
Context7	Upstash	Fast-moving libraries	Version-specific documentation on demand
feature-dev	Anthropic	Larger features	Exploration, architecture and review agents
frontend-design	Anthropic	UI work	Front-end output that avoids generic AI styling
LSP plugins	Anthropic	Typed codebases	Language-server diagnostics and navigation
Playwright	Microsoft	Browser testing	Browser automation and end-to-end tests
mcp-server-dev	Anthropic	Teams building MCP servers	Skills for server design, deployment and auth
skill-creator	Anthropic	Writing your own skills	Tools to create, test and measure skills

Best Claude Code Plugins And Skills In 2027

1. Superpowers — A Complete Engineering Method Packaged As Skills

Superpowers, maintained by Jesse Vincent, changes how Claude approaches work rather than what it can reach. It teaches brainstorming before building, subagent-driven development with built-in code review, systematic debugging and red/green test-driven development. It also teaches Claude how to write and test new skills of its own.

It is consistently among the most installed plugins in the official marketplace, and it pays off most on multi-step features where an unplanned agent tends to wander. Teams with an existing process may find parts of it opinionated, so try it on one project before adopting it everywhere.

2. code-review — Multi-Agent Pull Request Review That Filters False Positives

Anthropic’s code-review plugin runs several specialized agents over a pull request and scores each finding by confidence, dropping the ones that are likely noise. That scoring is the point: automated reviewers often bury real problems under trivial comments, and this one is built to report less but report what matters.

It suits teams that review many PRs and want a consistent first pass before a human looks. Pair it with pr-review-toolkit, another Anthropic plugin, if you want separate agents focused on tests, error handling and type design.

3. security-guidance — Security Checks On The Code Claude Writes

security-guidance reviews Claude’s own output for security problems. It raises pattern-based warnings as files are edited, runs a model-powered review of the diff when Claude stops, and adds a commit reviewer that looks for issues such as injection, cross-site scripting and server-side request forgery.

Agent-written code ships fast, and that is exactly when risky patterns slip through. Install it on any project that handles user input, and treat it as a guardrail alongside your existing scanners, not a replacement for them.

4. Context7 — Current Library Documentation Pulled In On Demand

Context7, from Upstash, gives Claude up-to-date, version-specific documentation and code examples for the libraries you use. The plugin connects to Context7’s hosted remote MCP server, so there is no local Node.js setup to maintain.

It fixes one of the most common agent failures: confidently calling an API that changed two releases ago. If your stack moves quickly, or you work across many libraries, this is usually the first integration worth adding.

5. feature-dev — A Full Feature Workflow With Specialist Agents

feature-dev is Anthropic’s structured workflow for building a whole feature. It brings agents for exploring the codebase, designing the architecture and reviewing the result, so Claude understands where a change fits before it starts editing files.

It is most useful in large or unfamiliar repositories, where the exploration step prevents changes that work in isolation but clash with existing patterns. For small fixes it adds overhead, so reserve it for real features.

6. frontend-design — Front-End Output That Does Not Look Generated

frontend-design steers Claude toward distinctive, production-grade interfaces instead of the generic look that many AI-generated pages share. It focuses on design quality, polish and deliberate visual choices rather than default component styling.

It has been one of the most installed plugins in the marketplace, which says a lot about how many developers ask agents for UI work. Use it on landing pages, dashboards and prototypes you plan to show to customers.

7. LSP Plugins — Language-Server Intelligence For Typed Codebases

Anthropic publishes language-server plugins for TypeScript, Python, Go, Rust, Java, Kotlin, Swift, C and C++, C#, PHP, Ruby and Lua. Once installed, Claude gets diagnostics after edits and can navigate by symbol instead of searching text, which cuts down on broken imports and type errors.

Each plugin needs the matching language server installed on your machine, such as the TypeScript or Pyright server. Install only the ones for the languages you actually use, since every plugin adds to what Claude loads.

8. Playwright — Browser Automation And End-To-End Testing

Microsoft’s Playwright plugin lets Claude drive a real browser: open pages, click, fill forms, take screenshots and run end-to-end tests. It is the most direct way to let an agent check that a front-end change actually works rather than assuming it does.

It is especially valuable alongside frontend-design or feature-dev, because Claude can verify the result in a browser before handing it back. Keep it scoped to test and staging environments rather than signed-in production accounts.

9. mcp-server-dev — Skills For Building MCP Servers That Work With Claude

mcp-server-dev is a set of Anthropic skills for designing and building MCP servers. It walks through deployment models, including remote HTTP servers, MCPB bundles and local servers, along with tool design patterns and authentication.

For teams exposing their own product or internal systems over MCP, it shortens the path from idea to a server that behaves well in Claude and other clients. Pair it with our list of MCP security tools before anything goes to production.

10. skill-creator — Write, Test And Measure Your Own Skills

skill-creator helps you build new skills, improve existing ones and measure how well they perform, including running evaluations and comparing versions. Skills are just folders of instructions and files, but writing ones that trigger reliably and actually improve results takes testing.

It is the right plugin once a team has repeatable work worth packaging, such as a release checklist or a house style for pull requests. Building one strong in-house skill often beats installing five generic ones.

How To Choose Claude Code Plugins

Which Gap Are You Closing?

Install plugins to fix a specific problem: missing documentation, weak reviews, untested UI or inconsistent process. A short set aimed at real gaps works better than a long list installed out of curiosity, because every plugin adds context Claude has to load.

Who Publishes It?

Plugins can run hooks and commands on your machine and connect to outside services. Prefer plugins published by Anthropic or by the vendor that owns the tool, check what each one will install before you accept it, and read our guide on how to secure AI agents and MCP integrations for the wider risks.

Do You Need A Plugin, A Skill Or Just An MCP Server?

A skill is a folder of instructions Claude loads when relevant. A plugin is the larger package that can combine skills, commands, hooks, subagents and MCP servers. If you only need a connection to a tool, adding its MCP server directly may be enough; our list of the best MCP servers covers the strongest options.

How Will You Keep Them Under Control?

Use /plugin list to see what is installed, /plugin disable to switch one off and /plugin uninstall to remove it. Review your set every few months, since plugins get renamed, merged or replaced as Claude Code evolves.

Conclusion

The best Claude Code setup is a small one. Start with a review plugin, security-guidance, Context7 and the language server for your main language, then add Superpowers or feature-dev if your team wants more structure. Playwright earns its place on any front-end project, and mcp-server-dev and skill-creator matter once you start building your own integrations and workflows. Install from the official marketplace, check what each plugin brings, and remove anything you stop using.