Complete Guide to Claude Code Plugins: From Official Plugins to Community Marketplaces

Complete Guide to Claude Code Plugins: From Official Plugins to Community Marketplaces

A deep dive into the Claude Code plugin system. Explore 13 official plugins in detail, 100+ community marketplace plugins, actual plugin.json structure and installation methods with practical examples.

Overview

Claude Code is an agentic coding tool developed by Anthropic. Through its plugin system, you can install slash commands, specialized agents, skills, and hooks as a single package.

This post provides an in-depth analysis of the 13 official plugins, over 100 community marketplace plugins, and the actual plugin structure.

★ Insight

  • 13 Official Plugins: Managed in the anthropics/claude-code repository
  • Community Marketplaces: 100+ plugins across 13 categories
  • Plugin Structure: commands/, agents/, skills/, hooks/, .mcp.json

Installing Plugins

Adding Marketplaces

Claude Code uses a Git repository-based marketplace system.

# Add community marketplaces
/plugin marketplace add ccplugins/awesome-claude-code-plugins
/plugin marketplace add kivilaid/plugin-marketplace
/plugin marketplace add ananddtyagi/cc-marketplace

# List registered marketplaces
/plugin marketplace list

Installing Plugins

# Basic installation
/plugin install commit-commands

# Install from a specific marketplace
/plugin install ultrathink@cc-marketplace

# List installed plugins
/plugin list

# Interactive installation (menu-driven)
/plugin

★ Insight

  • Minimum version requirement: Claude Code 1.0.33+
  • Storage location: Project level (.claude/) or user level (~/.claude/)
  • Development mode: Test locally with cc --plugin-dir /path/to/plugin

Deep Dive into 13 Official Plugins

Official plugins from the anthropics/claude-code repository managed by Anthropic.

1. commit-commands - Git Workflow Automation

/plugin install commit-commands

3 Commands Provided:

CommandFunction
/commitAnalyzes changes and auto-generates commit messages matching repository style
/commit-push-prBranch creation → Commit → Push → PR creation in one go
/clean_goneCleans up local branches deleted from remote, including worktrees

How /commit Works:

  1. Check changes with git status
  2. Analyze staged/unstaged changes with git diff
  3. Understand recent commit style with git log
  4. Generate commit message matching repository conventions
  5. Exclude sensitive files like .env, credentials.json

How /commit-push-pr Works:

  1. Automatically creates new branch if on main
  2. Analyzes entire commit history of branch (not just latest commit)
  3. Auto-generates PR description (Summary + Test plan)
  4. Creates PR via GitHub CLI (gh) and returns URL
# Usage example
git add .
/commit
# → Suggests "feat(auth): add JWT validation middleware"
# → Approve? [Y/n]

2. feature-dev - 7-Phase Feature Development Workflow

/plugin install feature-dev

7-Phase Structured Workflow:

graph TD
    A["1. Discovery<br/>Clarify requirements"] --> B["2. Codebase Exploration<br/>2-3 code-explorer agents in parallel"]
    B --> C["3. Clarifying Questions<br/>Edge cases, error handling"]
    C --> D["4. Architecture Design<br/>2-3 code-architect agents in parallel"]
    D --> E["5. Implementation<br/>Start after explicit approval"]
    E --> F["6. Quality Review<br/>3 code-reviewer agents in parallel"]
    F --> G["7. Summary<br/>Change summary and next steps"]

3 Specialized Agents:

AgentRole
code-explorerTraces execution paths, analyzes architecture layers, generates essential file lists
code-architectDesigns multiple approaches: minimal changes / clean architecture / pragmatic balance
code-reviewerFocused reviews on simplicity/DRY, bugs/correctness, conventions/abstractions
# Usage example
/feature-dev Add OAuth login with Google
# → Proceeds through 7 phases, user confirmation at each step

When to Use?

  • New features spanning multiple files
  • Complex tasks requiring architecture decisions
  • Features with unclear requirements

3. code-review - Confidence-Based Automated PR Review

/plugin install code-review

How It Works:

  1. Auto-skips closed/draft/trivial/already-reviewed PRs
  2. Collects CLAUDE.md guideline files
  3. Runs 4 parallel agents:
    • Agent #1, #2: CLAUDE.md compliance (duplicate verification)
    • Agent #3: Bug detection (within PR change scope only)
    • Agent #4: Git blame history analysis
  4. Assigns confidence score (0-100) to each issue
  5. Filters out scores below 80 (removes false positives)

Confidence Score Criteria:

ScoreMeaning
0No confidence, likely false positive
25Some confidence, might be real
50Moderate confidence, real but minor
75High confidence, real and important
100Absolute confidence, definitely a problem
# Terminal output
/code-review

# Post as PR comment
/code-review --comment

Filtered False Positives:

  • Issues that existed before the PR
  • Issues linters can catch
  • General quality issues (if not in CLAUDE.md)
  • Code with lint ignore comments

4. hookify - Create Hooks Without Coding

/plugin install hookify

4 Commands Provided:

CommandFunction
/hookify [description]Create hooks with natural language
/hookify (no args)Analyze conversation to auto-discover patterns to improve
/hookify:listList active rules
/hookify:configureEnable/disable rules

Hook File Format (.claude/hookify.*.local.md):

---
name: block-dangerous-rm
enabled: true
event: bash
pattern: rm\s+-rf
action: block  # warn or block
---

⚠️ **Dangerous rm command detected!**

This command could delete important files.
- Verify the path is correct
- Make sure you have backups

Event Types:

  • bash: On Bash command execution
  • file: On Edit/Write tool usage
  • stop: When Claude tries to end work
  • prompt: On user prompt submission
  • all: All events
# Usage example
/hookify Warn me when console.log is used in TypeScript files
# → Creates .claude/hookify.warn-console-log.local.md

5. plugin-dev - 8-Phase Plugin Development Tool

/plugin install plugin-dev

8-Phase Plugin Creation Workflow:

  1. Discovery: Understand plugin purpose and requirements
  2. Component Planning: Decide on skills/commands/agents/hooks/MCP needed
  3. Detailed Design: Detailed design for each component
  4. Structure Creation: Create directories and manifest
  5. Component Implementation: AI-assisted implementation
  6. Validation: plugin-validator and component-specific checks
  7. Testing: Real-world testing in Claude Code
  8. Documentation: Finalize README and prepare for distribution

7 Core Skills:

SkillTrigger KeywordsContent
Hook Development”create a hook”, “PreToolUse hook”Prompt/command hooks, all event types
MCP Integration”add MCP server”, “.mcp.json”stdio/SSE/HTTP/WebSocket server setup
Plugin Structure”plugin.json manifest”Standard directory structure, auto-discovery
Plugin Settings”.local.md files”Per-project settings, YAML frontmatter
Command Development”create a slash command”Command structure, argument handling
Agent Development”create an agent”Agent frontmatter, system prompts
Skill Development”create a skill”SKILL.md, progressive disclosure
# Usage example
/plugin-dev:create-plugin Database migration management plugin
# → 8-phase guide proceeds

6. frontend-design - Auto-Activating UI Design Skill

/plugin install frontend-design

A skill that auto-activates during frontend work.

Core Principles:

  • Bold aesthetic choices: Avoid generic AI styles
  • Distinctive typography and color palettes
  • High-impact animations and visual details
  • Context-aware implementation
# Auto-activation examples
"Create a dashboard for a music streaming app"
"Build a landing page for an AI security startup"
# → frontend-design skill auto-applies

7. pr-review-toolkit - 6 Specialized Review Agents

/plugin install pr-review-toolkit

6 Specialized Agents:

  • Comment reviewer
  • Test reviewer
  • Error handling reviewer
  • Type reviewer
  • Code quality reviewer
  • Comprehensive reviewer
/pr-review-toolkit:review-pr

8. security-guidance - Security Warning Hook

/plugin install security-guidance

9 Security Patterns Monitored:

  • Injection attacks
  • XSS (Cross-site scripting)
  • SQL injection
  • Command injection
  • Path traversal
  • Insecure deserialization
  • Sensitive data exposure
  • Weak cryptography
  • Authentication/authorization flaws

Real-time warnings before tool execution via PreToolUse hook.

Other Official Plugins

PluginDescription
agent-sdk-devClaude Agent SDK development kit (/new-sdk-app)
claude-opus-4-5-migrationAuto-migration to Opus 4.5
explanatory-output-styleEducational insights about implementation choices
learning-output-styleLearning mode requesting 5-10 line code contributions
ralph-wiggumSelf-referential AI loops for iterative development

Community Marketplace Analysis

Major Marketplaces

MarketplacePlugin CountFeatures
ccplugins/awesome-claude-code-plugins100+13 categories, largest scale
kivilaid/plugin-marketplace87+Strong in document processing tools
ananddtyagi/cc-marketplaceDynamic updatesLive database integration
ivan-magda/claude-code-marketplaceSwift-focusedSwift 6 migration, iOS development

Key Plugins by Category (ccplugins)

Workflow Orchestration (8):

  • ultrathink: Coordinator directing 4 specialized sub-agents
  • lyra: Master-level AI prompt optimization expert
  • problem-solver-specialist: Systematic problem solving

Code Quality & Testing (16):

  • bug-detective: Step-by-step systematic debugging
  • double-check: Dual verification of changes
  • test-writer-fixer: Test writing and fixing
  • unit-test-generator: Automatic unit test generation
  • database-performance-optimizer: DB performance optimization

Git Workflow (14):

  • fix-github-issue: Auto-fix based on GitHub issues
  • create-pr: PR creation automation
  • update-branch-name: Branch name convention enforcement
  • husky: Git hook setup

Documentation (8):

  • analyze-codebase: Comprehensive codebase analysis
  • changelog-generator: Auto-generate changelogs
  • update-claudemd: Auto-update CLAUDE.md
  • openapi-expert: OpenAPI spec expert

Security & Compliance (7):

  • audit: Codebase security audit
  • enterprise-security-reviewer: Enterprise-grade security review
  • legal-compliance-checker: Legal compliance verification
  • data-privacy-engineer: Data privacy

Development Engineering (15):

  • ai-engineer: AI feature development specialist
  • backend-architect: Backend architecture design
  • flutter-mobile-app-dev: Flutter mobile apps
  • react-native-dev: React Native development

kivilaid Marketplace Specialized Tools

Document Processing Suite:

  • xlsx: Excel generation with formulas and data visualization
  • docx: Word documents with change tracking and formatting
  • pptx: PowerPoint with layouts and templates
  • pdf: PDF extraction, merging, splitting, form processing

Creative & Design:

  • algorithmic-art: p5.js-based generative art
  • canvas-design: PNG/PDF visual art generation
  • slack-gif-creator: Animated GIFs for Slack

davila7 Toolkit (10):

  • Git workflow, Supabase, Next.js/Vercel
  • Testing, security audit, AI/ML, DevOps
  • Document generation, performance optimization, project management

Plugin Structure Details

Standard Directory Structure

my-plugin/
├── .claude-plugin/
│   └── plugin.json              # Plugin metadata (required)
├── commands/                    # Slash commands
│   ├── command-name.md
│   └── another-command.md
├── agents/                      # Autonomous agents
│   ├── agent-name.md
│   └── another-agent.md
├── skills/                      # Reusable skills
│   ├── skill-name/
│   │   ├── SKILL.md
│   │   ├── references/
│   │   ├── examples/
│   │   └── scripts/
├── hooks/                       # Event hooks
│   ├── hooks.json
│   └── my-hook.sh
├── .mcp.json                    # MCP server configuration (optional)
└── README.md

plugin.json Example

{
  "name": "my-team-plugin",
  "version": "1.0.0",
  "description": "Team development workflow plugin",
  "commands": ["commands/"],
  "agents": ["agents/"],
  "skills": ["skills/"],
  "hooks": ["hooks/"]
}

Command File Format (YAML frontmatter + Markdown)

---
description: "Performs code review"
argument-hint: "[--comment]"
allowed-tools:
  - Read
  - Grep
  - Bash
---

# Code Review Command

## Instructions
1. Analyze changes in current branch
2. Check CLAUDE.md guidelines
3. Report issues with confidence scores
...

Agent File Format

---
name: code-explorer
description: |
  An agent that deeply analyzes codebases.

  <example>
  User: "Trace how authentication flow works"
  → code-explorer is invoked
  </example>
model: sonnet
color: blue
tools:
  - Read
  - Grep
  - Glob
---

# System Prompt

You are a codebase exploration expert...

Skill SKILL.md Format

---
description: |
  Frontend UI design skill.

  <trigger_phrases>
  - "create a dashboard"
  - "landing page design"
  - "UI component generation"
  </trigger_phrases>
auto_invoke: true
---

# Frontend Design Skill

## Core Principles
1. Avoid generic AI styles
2. Bold aesthetic choices
...

.mcp.json MCP Server Configuration

{
  "mcpServers": {
    "postgresql": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres"],
      "env": {
        "DATABASE_URL": "${DATABASE_URL}"
      }
    },
    "browser": {
      "type": "sse",
      "url": "https://mcp.example.com/browser",
      "headers": {
        "Authorization": "Bearer ${API_KEY}"
      }
    }
  }
}

MCP Server Types:

  • stdio: Local process (most common)
  • sse: Server-Sent Events (hosting/OAuth)
  • http: REST API
  • websocket: Real-time bidirectional

Plugin Management Commands

# Install
/plugin install <name>
/plugin install <name>@<marketplace>

# Uninstall
/plugin uninstall <name>

# List
/plugin list

# Update
/plugin update <name>
/plugin update --all

# Marketplace management
/plugin marketplace add <owner/repo>
/plugin marketplace remove <name>
/plugin marketplace list

# Interactive menu
/plugin

Starting a New Project

/plugin install commit-commands     # Git workflow
/plugin install hookify             # Enforce team conventions
/plugin install security-guidance   # Security warnings

Focused Feature Development

/plugin install feature-dev         # 7-phase workflow
/plugin install frontend-design     # UI auto-skill (for frontend)

Automating Code Reviews

/plugin install code-review         # Confidence-based review
/plugin install pr-review-toolkit   # 6 specialized agents

Plugin Development

/plugin install plugin-dev          # 8-phase development guide

Important Considerations

Trust Verification

  • Official plugins: Managed in anthropics/claude-code repository
  • Community plugins: Requires direct code review
  • Plugins can access local filesystem

Version Requirements

# Requires Claude Code 1.0.33+
claude --version

MCP Server Authentication

# Set API keys via environment variables
export DATABASE_URL="postgresql://..."
export API_KEY="sk-..."

Summary

The Claude Code plugin system provides a rich ecosystem of 13 official plugins and 100+ community plugins.

<strong>Key Takeaways</strong>:

PurposeRecommended Plugin
Git Workflow Automationcommit-commands (official)
Feature Development Automationfeature-dev (official, 7-phase)
Code Review Automationcode-review (official, confidence-based)
Custom Rule Creationhookify (official, no-coding)
Security Warningssecurity-guidance (official)
Plugin Developmentplugin-dev (official, 8-phase)
UI Designfrontend-design (official, auto)
Document Processingxlsx, docx, pptx, pdf (kivilaid)
Comprehensive Orchestrationultrathink (ccplugins)

References

Read in Other Languages

Was this helpful?

Your support helps me create better content. Buy me a coffee! ☕

About the Author

JK

Kim Jangwook

Full-Stack Developer specializing in AI/LLM

Building AI agent systems, LLM applications, and automation solutions with 10+ years of web development experience. Sharing practical insights on Claude Code, MCP, and RAG systems.