Browse the full directory of Claude Code plugins — commands, agents, skills, MCP servers, and more.
Browse plugins →Eight Claude Code plugins ranked by 7-day install velocity in early June 2026, with what each does and how to pick between them.
The directory tracked 36,633 plugins by 16,679 authors as of June 8, 2026, with 358,699 components indexed across nine types. Install activity over the trailing 30 days reached 24,049 events, and 1,929 plugins entered the index this past week (versus 2,091 the week before — supply growth is decelerating slightly). This piece looks at eight plugins with measurable 7-day install velocity and explains what each one does so you can pick deliberately rather than by star count.
Star counts here are GitHub repo stars — many of these repos host content well beyond a single Claude plugin, so high stars do not equal plugin adoption. Install velocity is the more honest signal, and that drives the ordering below. Numbers come from a snapshot taken on June 8, 2026; live numbers may differ by the time you read this.
Eight plugins below, ordered by 7-day install count from the June 8, 2026 snapshot.
juliusbrussee-caveman ships agents, skills, and hooks. It compresses Claude Code session communication into an ultra-terse "caveman" style aimed at cutting token usage by roughly 75% while preserving technical accuracy. It also delegates complex tasks to subagents for code location, surgical editing, and diff review, and compresses memory files, git commits, and review comments into minimal-token formats. Repo stars: 64,415.
obra-superpowers-2 bundles skills and hooks that enforce a structured TDD workflow — parallel task execution, code review, and root-cause debugging. It guides brainstorming into validated specs, creates isolated git worktrees for feature branches, and runs verification checkpoints before merging or committing. Repo stars: 220,993.
pbakaus-impeccable is a skills-only plugin focused on designing and iterating production-grade frontend interfaces with real code — layout, typography, motion, accessibility, and design systems — while detecting anti-patterns. Repo stars: 34,933.
affaan-m-everything-claude-code is the broadest plugin on this list, packing commands, agents, skills, and hooks. The repo describes 64 specialized agents, 261 skills, and 84 commands spanning TDD, code review, security auditing, deployment, and automation across 30+ tech stacks, plus persistent context, autonomous loops, and MCP conventions. Repo stars: 209,640.
nextlevelbuilder-ui-ux-pro-max is a skills bundle covering 50+ UI/UX styles, color palettes, font pairings, guidelines, and charts for React, Next.js, Vue, Svelte, Tailwind, React Native, and Flutter. Repo stars: 79,973.
anthropics-frontend-design-plugins-frontend-design-2 is a skills bundle that generates production-grade frontend UIs — components, pages, or full apps — with distinctive typography, motion, and layouts that avoid generic styles, across HTML, CSS, JavaScript, React, or Vue. Repo stars: 20,139.
forrestchang-andrej-karpathy-skills is a skills bundle that codifies behavioral guidelines for LLM-assisted coding: simplicity, surgical changes, surfacing assumptions, and verifiable success criteria. Repo stars: 168,322.
xrf9268-hue-llm-council-plugin ships commands, agents, skills, and hooks. It runs Claude, OpenAI Codex, and Gemini in parallel for multi-model code reviews and synthesizes their opinions into a consensus verdict, with automated setup and cleanup. Repo stars: 17 — much lower than peers on this list, but the 74 installs in 7 days show real usage.
A few patterns separate this set into clusters.
Token-efficiency versus orchestration. caveman is explicitly about reducing tokens consumed per session. ecc and superpowers go the other direction — they spend more tokens to orchestrate richer multi-step workflows. If your bottleneck is per-session cost, caveman is the only entry in this batch that targets that. If your bottleneck is process discipline (TDD, code review gates, verification before merge), superpowers is the closest fit.
Frontend-design cluster. impeccable, ui-ux-pro-max, and frontend-design are all skills-only and all aimed at frontend UI. They overlap in scope but emphasize different things: impeccable on accessibility and anti-pattern detection, ui-ux-pro-max on a large catalog of styles and palettes across many frameworks, frontend-design on producing visually distinct (non-generic) interfaces. Installing two of these together is only worth it if you actually need both breadth and depth — pick one to start.
Multi-model review. llm-council-plugin is the only entry that runs Claude alongside Codex and Gemini and merges their verdicts. Useful if you want a non-Claude opinion on the diff before merging; not useful if you only have Claude configured.
Behavioral guardrails. karpathy-skills is the only entry whose pitch is purely about constraining how the model writes code — simplicity, surgical changes, assumption surfacing — rather than adding capabilities.
Star count is not adoption. karpathy-skills carries 168,322 repo stars and 103 installs in 7 days. ecc carries 209,640 repo stars and 326 installs. Stars reflect a parent project's general reach; installs measure plugin usage specifically. Use the install number, not the star number, to gauge actual traction.
A pragmatic ordering for picking from this list:
If you're new to Claude Code plugins, install one at a time and observe behavior before stacking. Component composition across the directory — 65,970 commands, 53,234 agents, 222,015 skills, 7,776 hooks, 8,813 MCP servers, 545 LSP servers, 194 output styles, 100 themes, and 52 monitors — shows that skills dominate by roughly an order of magnitude, which matches what most of this list ships.
Browse the full trending sort on /plugins to see live install velocity beyond this snapshot, or browse /plugins by component type.
This is a snapshot from June 8, 2026. Install velocity moves week to week, especially in the long tail where a single weekend's traffic can reorder the ranks. Two takeaways worth carrying forward: install count and star count tell different stories, and most velocity on this date is concentrated in a small number of skills-heavy plugins (caveman, superpowers, impeccable). For the current ordering, check /plugins?sort=trending directly.
Reduce Claude Code token consumption by ~75% using ultra-compressed communication and delegation to compressed-output subagents for code review, editing, and exploration.
Enforce a structured TDD workflow with parallel task execution, isolated git worktrees, and systematic code review. Guides root cause investigation before fixes, verifies tests and linters pass before claiming completion, and dispatches subagents for independent tasks like fixing multiple test failures simultaneously.
Audits and improves production frontend interfaces with real working code changes—covering UX review, visual hierarchy, accessibility, responsive behavior, theming, typography, layout, color, motion, micro-interactions, and design systems.
Eight Claude Code plugins worth evaluating, selected by install velocity, star count, and component coverage — from TDD enforcement to token-saving workflows.
The utilities category has 12,406 plugins — 39% of the Claude plugin directory. Here are 7 worth evaluating for token optimization, UI design, framework tooling, and persistent memory.
A data-driven guide to MCP server plugins for Claude Code, covering persistent memory, browser automation, live documentation, and all-in-one solutions from the 7,159 MCP servers in the ecosystem.
Turn Claude Code into a multi-agent engineering platform with 542 skills, 184 commands, and 134 agents for end-to-end feature development, code review, build fixing, testing, security auditing, and deployment across 40+ languages and frameworks.
Access 50+ UI/UX styles, color palettes, font pairings, guidelines, and charts to plan, build, review, and optimize web and mobile interfaces across React, Next.js, Vue, Svelte, Tailwind, React Native, and Flutter.
Make intentional visual design decisions for UI builds and redesigns — choose aesthetic direction, typography, and layout that avoid templated defaults.
Reduces common LLM coding mistakes by enforcing behavioral guidelines for simplicity, surgical changes, assumption surfacing, and verifiable success criteria
Run a three-phase LLM consensus protocol where Claude, OpenAI Codex, and Google Gemini collaborate on code reviews and technical problem-solving, then synthesize their opinions into a single Markdown verdict