jeremylongshore / claude-code-plugins-plus-skills

Claude Code Skills & Plugins Hub **The largest open-source marketplace for Claude Code plugins and agent skills.** 346 plugins, 1,900+ skills, 16 community contributors — validated, graded, and ready to install. Or use Claude's built-in command: **Browse the marketplace** | **Explore plugins** | **Download bundles** > **Killer Skill of the Week** — executive-assistant-skills: *"The AI skills that fully replaced a $4k/mo executive assistant"* by Martin Gontovnikas. Research meetings, draft emails, manage Todoist action items — the complete EA workflow. Grade: A. > > Previous picks: skill-creator, cursor-pack, crypto-portfolio-tracker. See all at tonsofskills.com. --- 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 ( )** replaced static plugins as the primary extension point. Skills auto-activate based on conversation context — no slash commands needed. This repo now contains 1,916 validated skills across 22 categories. • ** ** is the official path variable for referencing files within a skill directory. It replaced the deprecated syntax. All 1,916 skills in this repo use the current variable. • **Two-tier validation**: Standard tier (default) follows the Anthropic spec with no required fields. Enterprise tier ( ) applies the Intent Solutions 100-point grading rubric. CI auto-selects enterprise. • **Progressive Disclosure Architecture (PDA)** keeps skill files concise (target: 150 lines) and offloads detailed references to a subdirectory. Our 100-point grading rubric enforces this. • **Tool permissions** ( frontmatter) give each skill fine-grained access control. Scoped patterns like limit what system commands a skill can run. • **New frontmatter fields**: (run in subagent), (subagent type), (hide from slash menu), (autocomplete), (lifecycle events), and (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 or in frontmatter. This README reflects the current state of the ecosystem as of March 2026. --- Quick Start **Option 1: CLI (Recommended)** **Option 2: Claude Built-in Commands** > Already using an older install? Run and re-add with the command above to switch to the new slug. **Browse the catalog:** Visit **claudecodeplugins.io** or explore --- How Agent Skills Work Agent Skills are instruction files ( ) that teach Claude Code **when** and **how** to perform specific tasks. Unlike commands that require explicit triggers, skills activate automatically when Claude detects relevant conversation context. The Activation Flow What a Skill Looks Like The Numbers | Metric | Count | |--------|-------| | Total skills | 1,916 | | Plugins (marketplace) | 346 | | Plugin categories | 22 | | Contributors | 16 | | Average skill grade | B (89.9/100) | | Production ready (A+B) | 88.9% | --- 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** ( **Guides** (90+ pages) • Mental Model (5 min) • Architecture Deep Dive (15 min) • Build Your Own (30 min) • Debugging Tips (15 min) • Orchestration Pattern (60 min) **Interactive Tutorials** • Skills (5 notebooks) • Plugins (4 notebooks) • Orchestration (2 notebooks) **Reference Implementation** • 5-Phase Workflow • Phase Contracts & Agents • Verification Scripts --- 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 a…