Executive Summary
Claude Code's ecosystem spans five layers: skills (SKILL.md instruction folders), plugins (distributable packages), MCP servers (external tool connections via the Model Context Protocol), connectors (cloud-side remote integrations on claude.ai), and a powerful CLI with headless automation and an Agent SDK.
Five skills ship built-in. Anthropic publishes 17 official example skills and 13 demo plugins. An official plugin marketplace is auto-available with LSP for 11 languages and service integrations. The Agent Skills open standard has been adopted by 30+ tools. The MCP ecosystem has grown to 16,000+ servers with an official registry backed by Anthropic, GitHub, and Microsoft. Claude.ai offers 50+ pre-built cloud connectors. The CLI supports full non-interactive automation via -p mode, structured JSON output, budget controls, and a Claude Agent SDK for Python and TypeScript.
1. Built-in Skills (Ship with Claude Code)
These five skills are bundled with every Claude Code installation. They are prompt-based (not hardcoded commands) and can spawn agents, read files, and adapt to your codebase.
| Skill | Purpose | Key Capability |
|---|---|---|
/batch <instruction> | Orchestrate large-scale parallel codebase changes | Decomposes work into 5-30 units, spawns one agent per unit in isolated git worktrees, each opens a PR |
/claude-api | Load Claude API reference for your project's language | Covers Python, TypeScript, Java, Go, Ruby, C#, PHP, cURL. Auto-activates on anthropic imports |
/debug [description] | Troubleshoot current Claude Code session | Reads session debug log, optionally focused on a described issue |
/loop [interval] <prompt> | Run a prompt repeatedly on an interval | Useful for polling deployments, babysitting PRs, periodic checks |
/simplify [focus] | Review changed files for code quality | Spawns 3 parallel review agents, aggregates findings, applies fixes |
Source: code.claude.com/docs/en/skills [1]
2. Official Anthropic Skills (anthropics/skills Repository)
The anthropics/skills GitHub repository contains 17 reference skills. Install the marketplace:
/plugin marketplace add anthropics/skills
2.1 Document Creation & Editing
| Skill | Description | License |
|---|---|---|
docx | Word document creation and editing | Source-available |
pdf | PDF document creation and editing | Source-available |
pptx | PowerPoint presentation creation and editing | Source-available |
xlsx | Excel spreadsheet creation and editing | Source-available |
doc-coauthoring | Collaborative document editing workflows | Apache 2.0 |
2.2 Creative & Design
| Skill | Description |
|---|---|
algorithmic-art | Generate algorithmic/generative art |
canvas-design | Design work on canvas-based layouts |
frontend-design | Production-grade frontend interfaces with design philosophy |
slack-gif-creator | Create animated GIFs for Slack |
theme-factory | Generate design themes and color schemes |
web-artifacts-builder | Build interactive web artifacts |
2.3 Development & Technical
| Skill | Description |
|---|---|
claude-api | Claude API reference and usage patterns |
mcp-builder | Build MCP (Model Context Protocol) servers |
webapp-testing | Test web applications systematically |
skill-creator | Meta-skill: create new skills |
2.4 Enterprise & Communication
| Skill | Description |
|---|---|
brand-guidelines | Apply brand guidelines to content |
internal-comms | Draft internal communications |
Source: github.com/anthropics/skills [2]
3. Official Demo Plugins (anthropics/claude-code Repository)
13 demo plugins showcasing the plugin system. Add the marketplace first:
/plugin marketplace add anthropics/claude-code
| Plugin | Description | Value |
|---|---|---|
code-review | Automated PR review with 5 parallel Sonnet agents for compliance, bugs, history, and comments | High |
feature-dev | 7-phase feature development workflow with agents for exploration, architecture, and quality | High |
commit-commands | Git workflow automation: /commit, /commit-push-pr, /clean_gone | High |
pr-review-toolkit | Specialized PR review agents for tests, error handling, type design, code quality | High |
security-guidance | Security reminder hook monitoring 9 patterns (XSS, injection, eval, pickle, etc.) | High |
plugin-dev | 8-phase guided workflow for building plugins, with 7 expert skills | Medium |
hookify | Custom hook creation system to prevent unwanted behaviors | Medium |
frontend-design | Production-grade frontend creation with design system guidance | Medium |
agent-sdk-dev | Development kit for Claude Agent SDK with /new-sdk-app command | Medium |
claude-opus-4-5-migration | Automated migration from Sonnet 4.x/Opus 4.1 to Opus 4.5 | Situational |
explanatory-output-style | Hook injecting implementation insights at session start | Low |
learning-output-style | Interactive learning mode requesting code contributions at decision points | Low |
ralph-wiggum | Autonomous iteration loops for self-referential development | Niche |
Source: github.com/anthropics/claude-code/tree/main/plugins [3]
4. Official Plugin Marketplace (claude-plugins-official)
Auto-available in every Claude Code install. Browse via /plugin > Discover tab.
4.1 Code Intelligence (LSP Plugins)
Enable jump-to-definition, find-references, and real-time type error detection. Require the language server binary on your system.
| Language | Plugin | Binary Required |
|---|---|---|
| C/C++ | clangd-lsp | clangd |
| C# | csharp-lsp | csharp-ls |
| Go | gopls-lsp | gopls |
| Java | jdtls-lsp | jdtls |
| Kotlin | kotlin-lsp | kotlin-language-server |
| Lua | lua-lsp | lua-language-server |
| PHP | php-lsp | intelephense |
| Python | pyright-lsp | pyright-langserver |
| Rust | rust-analyzer-lsp | rust-analyzer |
| Swift | swift-lsp | sourcekit-lsp |
| TypeScript | typescript-lsp | typescript-language-server |
4.2 External Integrations
| Category | Plugins |
|---|---|
| Source Control | github, gitlab |
| Project Management | atlassian (Jira/Confluence), asana, linear, notion |
| Design | figma |
| Infrastructure | vercel, firebase, supabase |
| Communication | slack |
| Monitoring | sentry |
4.3 Development Workflows & Output Styles
| Plugin | Description |
|---|---|
commit-commands | Git commit, push, and PR creation workflows |
pr-review-toolkit | Specialized PR review agents |
agent-sdk-dev | Claude Agent SDK development tools |
plugin-dev | Plugin creation toolkit |
explanatory-output-style | Educational insights about implementation choices |
learning-output-style | Interactive learning mode |
Sources: code.claude.com/docs/en/discover-plugins [4], github.com/anthropics/claude-plugins-official [5]
5. Agent Skills Open Standard
Anthropic released Agent Skills as an open standard on December 18, 2025. The specification lives at agentskills.io and github.com/agentskills/agentskills.
5.1 Adopting Tools (30+)
| Tool | Vendor |
|---|---|
| Claude Code / Claude.ai | Anthropic |
| Cursor | Anysphere |
| VS Code / GitHub Copilot | Microsoft / GitHub |
| Gemini CLI | |
| OpenAI Codex CLI | OpenAI |
| JetBrains Junie | JetBrains |
| Kiro | AWS |
| Roo Code | Community |
| Goose | Block |
| OpenHands | All Hands AI |
| Amp | Sourcegraph |
| Antigravity IDE | Antigravity |
5.2 SKILL.md Format
| Field | Required | Description |
|---|---|---|
name | Yes | Max 64 chars, lowercase + hyphens, must match directory name |
description | Yes | Max 1024 chars, what the skill does and when to use it |
license | No | License name or reference |
compatibility | No | Max 500 chars, environment requirements |
metadata | No | Arbitrary key-value pairs (author, version, etc.) |
allowed-tools | No | Space-delimited pre-approved tools (experimental) |
5.3 Claude Code Extensions (Beyond Standard)
| Field | Purpose |
|---|---|
disable-model-invocation | Prevent auto-loading (manual /name only) |
user-invocable | Set false to hide from / menu |
model | Override model for this skill |
effort | Override effort level (low, medium, high, max) |
context | Set to fork for isolated subagent execution |
agent | Subagent type when context: fork |
argument-hint | Autocomplete hint for arguments |
hooks | Hooks scoped to this skill's lifecycle |
Source: agentskills.io/specification [7]
6. Community Skills Catalogs
6.1 VoltAgent/awesome-agent-skills (630+ skills)
The largest curated catalog, organized by publisher.
| Publisher | # Skills | Highlights |
|---|---|---|
| Vercel | 8 | react-best-practices, next-best-practices, web-design-guidelines, composition-patterns |
| Cloudflare | 7 | agents-sdk, building-ai-agent, durable-objects, web-perf, wrangler |
| Netlify | 12 | functions, edge-functions, blobs, db, image-cdn, forms, frameworks |
| 4 | gemini-api-dev, vertex-ai-api-dev, gemini-live-api-dev | |
| Google Workspace CLI | 12 | gws-drive, gws-sheets, gws-gmail, gws-calendar, gws-docs, gws-slides |
| Stripe | 2 | stripe-best-practices, upgrade-stripe |
| HashiCorp | 3 | terraform-code-generation, terraform-module-generation |
| Firecrawl | 8 | cli, agent, browser, crawl, download, map, scrape, search |
| Better Auth | 7 | best-practices, providers, create-auth, organization, twoFactor |
| Supabase | 1 | postgres-best-practices |
| Neon | 3 | neon-postgres, claimable-postgres |
| Sanity | 4 | best-practices, content-modeling, seo-aeo-best-practices |
| Tinybird | 4 | best-practices, cli-guidelines, python-sdk, typescript-sdk |
| Composio | 1000+ | Managed authentication for 1000+ external app integrations |
Source: github.com/VoltAgent/awesome-agent-skills [9]
6.2 obra/superpowers (13 workflow skills)
| Skill | Description |
|---|---|
test-driven-development | TDD workflow enforcement |
systematic-debugging | Structured debugging methodology |
verification-before-completion | Verify work before marking done |
brainstorming | Structured brainstorming sessions |
writing-plans | Plan authoring workflow |
executing-plans | Plan execution tracking |
dispatching-parallel-agents | Multi-agent orchestration |
requesting-code-review | Code review request workflow |
receiving-code-review | Code review response workflow |
using-git-worktrees | Git worktree management |
finishing-a-development-branch | Branch completion checklist |
subagent-driven-development | Development via subagent delegation |
writing-skills | Meta-skill for creating skills |
Source: github.com/obra/superpowers [10]
6.3 Third-Party Directories
| Site | URL | Description |
|---|---|---|
| SkillsMP | skillsmp.com | Aggregator claiming 87K+ skills from GitHub repos |
| Claude Skills Info | claudeskills.info | 658+ skills across 13 categories |
| Awesome Skills | awesome-skills.com | Visual skills directory |
| Claude Plugins Dev | claude-plugins.dev | Community plugin registry with CLI |
| Awesome Claude | awesomeclaude.ai | Visual directory with 40 auto-activating skills |
7. MCP Servers & Connectors
MCP (Model Context Protocol) is an open standard created by Anthropic for connecting AI tools to external systems. Now governed by the Linux Foundation, supported by Claude, ChatGPT, VS Code Copilot, Cursor, and 50+ clients. The ecosystem has grown to 16,000+ servers.
7.1 How MCP Works in Claude Code
| Transport | Command | Use Case |
|---|---|---|
| HTTP (recommended) | claude mcp add --transport http <name> <url> | Remote cloud services |
| stdio | claude mcp add <name> -- <command> [args] | Local processes on your machine |
| SSE (deprecated) | claude mcp add --transport sse <name> <url> | Legacy, replaced by HTTP |
7.2 Configuration Scopes
| Scope | Storage | Visibility |
|---|---|---|
| Local (default) | ~/.claude.json | Private, current project only |
| Project | .mcp.json at project root | Shared via version control |
| User | ~/.claude.json | All your projects |
| Managed | managed-mcp.json | Admin-deployed, organization-wide |
7.3 MCP Registries
| Registry | Scale | Notes |
|---|---|---|
| Official MCP Registry | Preview | Backed by Anthropic, GitHub, Microsoft, PulseMCP |
| Anthropic Subregistry | Curated | Powers Claude Code's server discovery |
| Glama | 14,274 servers | Largest community directory |
| PulseMCP | 7,600+ servers | With newsletter and blog |
| GitHub MCP Registry | 87 servers | Namespace-verified |
| mcp.so | Community | Community-curated |
7.4 Claude.ai Connectors (Remote MCP)
50+ pre-built cloud connectors at claude.ai/settings/connectors. Available across claude.ai, Co-Work, and Claude Code.
| Category | Connectors |
|---|---|
| Communication | Gmail, Slack, Intercom |
| Project Management | Google Calendar, Asana, Linear, Atlassian, Notion, Airtable |
| Design | Figma, Canva |
| Engineering | GitHub, Hex |
| Business | Stripe, Amplitude, Ahrefs, Clay, WordPress.com, Salesforce |
| Automation | Zapier (8,000+ connected apps) |
| Storage | Google Drive |
Custom connectors: On paid plans, enter any remote MCP server URL at Settings > Connectors > "Add custom connector." Free users limited to one.
7.5 Popular MCP Servers for Developers
| Server | What It Does |
|---|---|
| GitHub | Repository management, PR workflows, issues |
| PostgreSQL / Supabase | Direct database access and queries |
| Playwright | Browser automation and testing |
| Sentry | Production error monitoring |
| Notion | Documentation and wiki access |
| Composio | Managed gateway to 500+ apps with OAuth |
| Context7 | Live library documentation lookup |
| Memory | Persistent memory across sessions |
7.6 MCP SDKs
The github.com/modelcontextprotocol organization has 39 repositories with SDKs for TypeScript (11.9K stars), Python, C# (4.1K), Rust (3.2K), Java, Kotlin, Go, PHP, Ruby, and Swift. The Inspector tool (9.2K stars) provides visual testing for MCP servers.
Sources: code.claude.com/docs/en/mcp [16], modelcontextprotocol.io [18]
8. CLI & Programmatic Access
Claude Code's CLI supports full headless automation, structured output, and a programmatic SDK.
8.1 Core CLI Commands
| Command | Description |
|---|---|
claude | Start interactive session |
claude "query" | Start interactive with initial prompt |
claude -p "query" | Non-interactive print mode — run and exit |
cat file | claude -p "query" | Pipe content into print mode |
claude -c | Continue most recent conversation |
claude -r "<session>" "query" | Resume session by ID or name |
claude update | Update to latest version |
claude mcp | Configure MCP servers |
claude agents | List configured subagents |
claude remote-control | Start Remote Control server |
8.2 Headless Mode Key Flags
| Flag | Purpose |
|---|---|
--output-format text|json|stream-json | Control output format |
--json-schema '{...}' | Get validated structured output |
--max-turns N | Limit agentic turns |
--max-budget-usd N | Maximum dollar spend |
--allowedTools "Read,Edit,Bash" | Auto-approve specific tools |
--model sonnet|opus | Set model |
--effort low|medium|high|max | Set effort level |
--worktree | Run in isolated git worktree |
--fallback-model | Auto-fallback when overloaded |
8.3 Permission Modes for Automation
| Mode | Behavior |
|---|---|
--permission-mode default | Ask user for each tool use |
--permission-mode acceptEdits | Auto-approve file edits |
--permission-mode plan | Read-only, no modifications |
--permission-mode dontAsk | Auto-deny unless in allowedTools |
--dangerously-skip-permissions | Skip all prompts (containers only) |
--permission-prompt-tool | Delegate to MCP tool |
8.4 Claude Agent SDK
Formerly "Claude Code SDK." Provides the same agent loop, tools, and context management as Claude Code, usable programmatically.
| Language | Install |
|---|---|
| Python | pip install claude-agent-sdk |
| TypeScript | npm install @anthropic-ai/claude-agent-sdk |
Core API: The
query() function gives you the full Claude Code agent loop as a library. Supports hooks, subagents, MCP servers, permissions, sessions, skills, and plugins.8.5 GitHub Actions Integration
Official action: anthropics/claude-code-action@v1. Supports @claude mention triggers in PR comments, AWS Bedrock, and Google Vertex AI backends. GitLab CI/CD also documented.
8.6 Notable Advanced Flags
| Flag | Purpose |
|---|---|
--remote | Create web session on claude.ai |
--teleport | Resume web session in local terminal |
--remote-control | External control server |
--chrome / --no-chrome | Enable/disable Chrome integration |
--agents | Define subagents via inline JSON |
--plugin-dir | Load plugins from directory |
--from-pr | Resume sessions linked to a PR |
--teammate-mode | Agent display: auto, in-process, tmux |
Sources: code.claude.com/docs/en/cli-reference [21], code.claude.com/docs/en/headless [22]
9. Curated Recommendations
Based on profile: 20+ years experience, Django/Python, Windows, Google Workspace, ClickUp, Stripe, Cloudflare already connected.
7.1 Install Immediately
| What | Source | Why |
|---|---|---|
pyright-lsp | Official marketplace | Real-time Python type checking after every edit — critical for Django |
code-review | Demo marketplace | 5 parallel agents review PRs for bugs, compliance, and patterns |
commit-commands | Demo/Official | Streamlined git workflow with /commit and /commit-push-pr |
security-guidance | Demo marketplace | Automatic security pattern monitoring |
docx, pdf, xlsx | Anthropic skills | Create/edit Office documents and PDFs directly |
skill-creator | Anthropic skills | Meta-skill to build new skills faster |
7.2 Install When Needed
| What | Source | Why |
|---|---|---|
feature-dev | Demo marketplace | Structured 7-phase feature development workflow |
hookify | Demo marketplace | Create custom hooks to prevent unwanted behaviors |
mcp-builder | Anthropic skills | Build custom MCP servers |
webapp-testing | Anthropic skills | Systematic web application testing |
stripe-best-practices | VoltAgent catalog | Complements your Stripe MCP connection |
obra/superpowers | Community | Development discipline: TDD, debugging, verification |
7.3 Quick Start Commands
# Check version (need 1.0.33+)
claude --version
# Browse official marketplace (auto-available)
/plugin
# Add Anthropic skills marketplace
/plugin marketplace add anthropics/skills
# Add demo plugins marketplace
/plugin marketplace add anthropics/claude-code
# Install priority plugins
/plugin install pyright-lsp@claude-plugins-official
/plugin install code-review@anthropics-claude-code
/plugin install commit-commands@anthropics-claude-code
/plugin install security-guidance@anthropics-claude-code
# Reload
/reload-plugins
10. Architecture: Full Ecosystem Map
| Concept | What It Is | Distribution |
|---|---|---|
| Skill | A folder with SKILL.md + optional files | Personal (~/.claude/skills/), Project (.claude/skills/), or inside a plugin |
| Plugin | A package with .claude-plugin/plugin.json manifest | Via marketplaces, git repos, or local paths |
| Marketplace | A catalog of plugins | GitHub repos, git URLs, local paths, or remote URLs |
| MCP Server (local) | External tool running as a local process | claude mcp add or .mcp.json |
| MCP Server (remote) | Cloud service accessed via HTTP | URL + OAuth authentication |
| Connector | Pre-built cloud integration on claude.ai | claude.ai/settings/connectors |
| Agent SDK | Programmatic library (query() function) | pip install claude-agent-sdk |
CLI (-p mode) | Non-interactive automation | claude -p "prompt" with flags |
| Custom Command | Legacy .claude/commands/*.md file | Merged into skills system |
| Built-in Command | Hardcoded CLI operations (/clear, /compact, /help) | Ships with Claude Code |
| Built-in Skill | Prompt-based bundled skills (/batch, /simplify) | Ships with Claude Code |
Priority order: Enterprise skills > Personal skills > Project skills. Plugin skills are namespaced (plugin-name:skill-name) and cannot conflict. Custom commands have been merged into skills — existing .claude/commands/ files keep working.
11. Limitations & Caveats
- Community catalog numbers are unverified. SkillsMP claims 87K+ skills; these aggregate from public GitHub repos with minimal curation. Quality varies widely.
- Plugin system is in public beta. Launched October 9, 2025. Functional but still evolving.
- Security is user-responsible. Anthropic warns: "Anthropic does not control what MCP servers, files, or other software are included in plugins." [4]
- Cross-tool portability has limits. Claude Code's extended frontmatter fields (context: fork, agent, hooks) are Claude Code-specific.
- LSP plugins require system binaries. Installing pyright-lsp requires pyright-langserver installed separately.
- Skill context budget. Many skills may exceed the character budget (2% of context window, fallback 16K chars). Check with /context.
Bibliography
[1] Anthropic. "Extend Claude with skills." Claude Code Docs. code.claude.com/docs/en/skills
[2] Anthropic. "Public repository for Agent Skills." GitHub. github.com/anthropics/skills
[3] Anthropic. "Claude Code Plugins." GitHub. github.com/anthropics/claude-code/tree/main/plugins
[4] Anthropic. "Discover and install prebuilt plugins." Claude Code Docs. code.claude.com/docs/en/discover-plugins
[5] Anthropic. "Official Claude Code Plugins." GitHub. github.com/anthropics/claude-plugins-official
[6] SiliconANGLE. "Anthropic Makes Agent Skills Open Standard." Dec 18, 2025. siliconangle.com
[7] Agent Skills. "Specification." agentskills.io/specification
[8] Agent Skills. "Specification and documentation." GitHub. github.com/agentskills/agentskills
[9] VoltAgent. "Awesome Agent Skills." GitHub. github.com/VoltAgent/awesome-agent-skills
[10] obra. "Superpowers." GitHub. github.com/obra/superpowers
[11] travisvn. "Awesome Claude Skills." GitHub. github.com/travisvn/awesome-claude-skills
[12] hesreallyhim. "Awesome Claude Code." GitHub. github.com/hesreallyhim/awesome-claude-code
[13] rohitg00. "Awesome Claude Code Toolkit." GitHub. github.com/rohitg00/awesome-claude-code-toolkit
[14] alirezarezvani. "Claude Skills." GitHub. github.com/alirezarezvani/claude-skills
[15] ccplugins. "Awesome Claude Code Plugins." GitHub. github.com/ccplugins/awesome-claude-code-plugins
[16] Anthropic. "Connect Claude Code to tools via MCP." Claude Code Docs. code.claude.com/docs/en/mcp
[17] Model Context Protocol. "Official MCP Registry." registry.modelcontextprotocol.io
[18] Model Context Protocol. "MCP Official Site." modelcontextprotocol.io
[19] Anthropic. "Pre-built web connectors using remote MCP." support.claude.com
[20] Model Context Protocol. "GitHub Organization." github.com/modelcontextprotocol
[21] Anthropic. "CLI Reference." Claude Code Docs. code.claude.com/docs/en/cli-reference
[22] Anthropic. "Run Claude Code programmatically." Claude Code Docs. code.claude.com/docs/en/headless
[23] Anthropic. "Claude Agent SDK Overview." platform.claude.com/docs/en/agent-sdk/overview
[24] Anthropic. "GitHub Actions." Claude Code Docs. code.claude.com/docs/en/github-actions