Skip to content

README.md

Latest commit

 

History

History
362 lines (263 loc) · 20.4 KB

README.md

File metadata and controls

362 lines (263 loc) · 20.4 KB

Claude Code Skills & Plugins Hub

Version CLI Agent Skills Plugins 2026 Schema Tool Permissions Interactive Tutorials Open In Colab GitHub Stars Powered by Anthropic Built on Google Cloud Buy me a monster $5/month Sponsor

The largest open-source marketplace for Claude Code plugins and agent skills. 347 plugins, 1,900+ skills, 69 community contributions — validated, graded, and ready to install.

pnpm add -g @intentsolutionsio/ccpi    # Install the CLI
ccpi install devops-automation-pack     # Install any plugin

Or use Claude's built-in command:

/plugin marketplace add jeremylongshore/claude-code-plugins

Browse the marketplace | Explore plugins | Download bundles

Killer Skill of the Week — Every week we scour GitHub for the most impressive Claude Code skills and spotlight one that's popping. Sign up at tonsofskills.com and we'll email you the pick — no searching required.

What Changed: Claude Code in 2026

Anthropic shipped several features since plugins launched in October 2025 that change how developers extend Claude Code:

  • Agent Skills (SKILL.md) replaced static plugins as the primary extension point. Skills auto-activate based on conversation context — no slash commands needed. This repo now contains 1,900+ validated skills across 22 categories.
  • ${CLAUDE_SKILL_DIR} is the official path variable for referencing files within a skill directory. It replaced the deprecated {baseDir} syntax. All 1,900+ skills in this repo use the current variable.
  • Progressive Disclosure Architecture (PDA) keeps skill files concise (target: 150 lines) and offloads detailed references to a references/ subdirectory. Our 100-point grading rubric enforces this.
  • Tool permissions (allowed-tools frontmatter) give each skill fine-grained access control. Scoped patterns like Bash(git:*) limit what system commands a skill can run.
  • New frontmatter fields: context: fork (run in subagent), agent (subagent type), user-invocable (hide from slash menu), argument-hint (autocomplete), hooks (lifecycle events), and compatibility (AgentSkills.io standard for environment requirements).
  • Claude 4.5/4.6 model family — Opus 4.6, Sonnet 4.6, Haiku 4.5 are the current models. Skills can specify model: sonnet or model: opus in frontmatter.

This README reflects the current state of the ecosystem as of March 2026.

Quick Start

Option 1: CLI (Recommended)

pnpm add -g @intentsolutionsio/ccpi
ccpi search devops              # Find plugins by keyword
ccpi install devops-automation-pack
ccpi list --installed           # See what's installed
ccpi update                     # Pull latest versions

Option 2: Claude Built-in Commands

/plugin marketplace add jeremylongshore/claude-code-plugins
/plugin install devops-automation-pack@claude-code-plugins-plus

Already using an older install? Run /plugin marketplace remove claude-code-plugins and re-add with the command above to switch to the new slug.

Browse the catalog: Visit claudecodeplugins.io or explore plugins/

How Agent Skills Work

Agent Skills are instruction files (SKILL.md) that teach Claude Code when and how to perform specific tasks. Unlike commands that require explicit /slash triggers, skills activate automatically when Claude detects relevant conversation context.

The Activation Flow

1. INSTALL     /plugin install ansible-playbook-creator@claude-code-plugins-plus
2. STARTUP     Claude reads SKILL.md frontmatter → learns trigger phrases
3. ACTIVATE    You say "create an ansible playbook for Apache"
4. EXECUTE     Claude matches trigger → reads full skill → follows instructions

What a Skill Looks Like

---
name: ansible-playbook-creator
description: |
  Generate production-ready Ansible playbooks. Use when automating server
  configuration or deployments. Trigger with "ansible playbook" or
  "create playbook for [task]".
allowed-tools: Read, Write, Bash(ansible:*), Glob
version: 2.0.0
author: Jeremy Longshore <jeremy@intentsolutions.io>
license: MIT
---

# Ansible Playbook Creator

## Overview
Generates idempotent Ansible playbooks following infrastructure-as-code best practices.

## Instructions
1. Gather target host details and desired state
2. Select appropriate Ansible modules
3. Generate playbook with proper variable templating
4. Validate syntax with `ansible-lint`

## Output
- Complete playbook YAML ready for `ansible-playbook` execution

The Numbers

Metric Count
Total skills 1,900+
Plugin categories 22
Community contributions 69 plugins
Average skill grade B (87/100)
Production ready (A+B) 80%+

Plugin Types

AI Instruction Plugins (98%)

Markdown files that guide Claude's behavior through structured instructions, skills, commands, and agents. No external code — everything runs through Claude's built-in capabilities.

MCP Server Plugins (2%)

TypeScript applications that run as separate Node.js processes. Claude communicates with them through the Model Context Protocol. Currently 7 MCP plugins providing 21 tools.

SaaS Skill Packs

Pre-built skill collections for specific platforms — Deepgram, LangChain, Linear, Gamma, and 20+ others. Each pack includes install/auth, core workflows, debugging, deployment, and advanced pattern skills.

Skill Quality: 100-Point Grading

Every skill in this marketplace is automatically graded on a 100-point scale across five dimensions:

Dimension Weight What It Measures
Progressive Disclosure 30 pts File length, references/ usage, navigation
Ease of Use 25 pts Metadata quality, discoverability, workflow clarity
Utility 20 pts Problem solving, flexibility, error handling, examples
Spec Compliance 15 pts Frontmatter validity, naming, description quality
Writing Style 10 pts Imperative voice, objectivity, conciseness

Grades: A (90-100), B (80-89), C (70-79), D (60-69), F (<60)

Run the validator yourself:

python3 scripts/validate-skills-schema.py --skills-only          # Grade all skills
python3 scripts/validate-skills-schema.py --verbose --skills-only # Detailed breakdowns
python3 scripts/validate-skills-schema.py --show-low-grades       # Find D+F skills

Building Your Own

Templates

Template Includes Best For
minimal-plugin plugin.json + README Simple utilities
command-plugin Slash commands Custom workflows
agent-plugin Specialized AI agent Domain expertise
full-plugin Commands + agents + hooks + skills Complex automation

All templates live in templates/.

Step by Step

  1. Copy a template: cp -r templates/command-plugin my-plugin
  2. Edit .claude-plugin/plugin.json with your metadata
  3. Add your skill to skills/my-skill/SKILL.md
  4. Validate: ccpi validate ./my-plugin
  5. Submit a pull request

Skill Frontmatter Reference

---
# Required
name: my-skill-name                     # kebab-case, matches folder name
description: |                          # Include "Use when..." and "Trigger with..."
  Describe what this skill does. Use when building X.
  Trigger with "build X" or "create X workflow".
allowed-tools: Read, Write, Bash(npm:*) # Comma-separated, scoped Bash recommended
version: 1.0.0                          # Semver
author: Your Name <you@example.com>
license: MIT

# Optional
model: sonnet                           # Model override (sonnet, haiku, opus)
context: fork                           # Run in subagent
agent: Explore                          # Subagent type
user-invocable: false                   # Hide from / menu
argument-hint: "<file-path>"            # Autocomplete hint
hooks: {}                               # Lifecycle hooks
compatibility: "Node.js >= 18"          # Environment requirements
compatible-with: claude-code, cursor    # Platform compatibility
tags: [devops, ci]                      # Discovery tags
---

Path variable: Use ${CLAUDE_SKILL_DIR} to reference files relative to the skill directory.

Learning Lab

Production agent workflow patterns with empirical verification — guides, diagrams, and working examples.

Start Here (5 min) | Architecture Map | System Summary

Guides (90+ pages)

Interactive Tutorials Open In Colab

Reference Implementation

Contributors

Community contributors make this marketplace better. Newest first.

  • @mjaskolski (Michal Jaskolski) — Added 25 externally-synced agent skills from wondelai/skills covering product strategy, UX design, marketing/CRO, sales/influence, and growth frameworks. (#303)
  • @clowreed (B12.io) — Created b12-claude-plugin, an official B12.io plugin with a website-generator skill that takes users from idea to production-ready website draft in one conversation. (#307)
  • @duskfallcrew (Duskfall Crew) — Reported PHP webshell payloads in penetration-tester plugin README that triggered AV false positives. Drove a complete v2.0.0 rebuild with real Python security scanners. (#300)
  • @rowanbrooks100 (Rowan Brooks) — Created brand-strategy-framework, a 7-part brand strategy methodology used by top agencies with Fortune 500 clients. (#292)
  • @RichardHightower (Rick Hightower) — Creator of SkillzWave (44,000+ agentic skills). His quality reviews exposed validation gaps and drove 4,300+ lines of fixes plus new validator checks. Author of the Claude Code Skills Deep Dive series. (#293, #294, #295)
  • @TomLucidor (Tom) — His question about paid API requirements sparked the "Make All Plugins Free" initiative (v4.1.0) and drove 2,400+ lines of constraint documentation across 6 plugins. (#148)
  • @alexfazio (Alex Fazio) — Production agent workflow patterns and validation techniques that inspired the Learning Lab system (v4.0.0).
  • @lukeslp (Lucas Steuber) — Created geepers-agents with 51 specialized agents for development, deployment, quality audits, research, and game development. (#159)
  • @BayramAnnakov (Bayram Annakov) — Created claude-reflect, a self-learning system that captures corrections and syncs them to CLAUDE.md. (#241)
  • @jleonelion (James Leone) — Fixed bash variable scoping bug in Learning Lab scripts and improved markdown formatting. (#239)
  • @CharlesWiltgen (Charles Wiltgen) — Created Axiom, iOS development plugin with 13 production-ready skills for Swift/Xcode. (#121)
  • @aledlie (Alyshia Ledlie) — Fixed 7 critical JSON syntax errors and added production CI/CD patterns. (#117)
  • @JackReis (Jack Reis) — Contributed neurodivergent-visual-org plugin with ADHD-friendly Mermaid diagrams. (#106)
  • @terrylica (Terry Li) — Built prettier-markdown-hook with zero-config markdown formatting. (#101)
  • @beepsoft — Quality feedback on skill implementations that drove ecosystem-wide improvements. (#134)
  • @clickmediapropy — Reported mobile horizontal scrolling bug. (#120)

Want to contribute? See CONTRIBUTING.md or reach out to jeremy@intentsolutions.io

Resources

Official Anthropic

Technical Deep Dives

Community

Ecosystem

Documentation & Playbooks (click to expand)
Document Purpose
User Security Guide How to safely evaluate and install plugins
Code of Conduct Community standards
Security Policy Threat model, vulnerability reporting
Changelog Release history

Production Playbooks (11 guides, ~53,500 words):

Important Notes

Not on GitHub Marketplace. Claude Code plugins use a separate ecosystem with JSON-based catalogs hosted in Git repositories. This repository is a Claude Code plugin marketplace.

Free and open-source. All plugins are MIT-licensed. No monetization mechanism exists for Claude Code plugins. See Monetization Alternatives for external revenue strategies.

Production status. Claude Code plugins launched in public beta (October 2025) and are now a stable part of the Claude Code workflow. This marketplace tracks all specification changes.

License

MIT License — See LICENSE for details.

Acknowledgments

  • Anthropic — For Claude Code and the plugin/skills system
  • Richard Hightower — Co-founder of SkillzWave; his quality reviews shaped our grading rubric
  • Lee-Han Chung — Definitive Claude Skills Deep Dive technical analysis
  • Community Contributors — Everyone who builds, reviews, and improves the ecosystem

Version: 4.17.0 | Last Updated: March 2026 | Skills: 1,900+ | Plugins: 347