# Agent Skill Hub > Open-source Agent Skills, Claude Code and Cursor plugins, MCP servers, agent harnesses and CLIs for AI coding agents. Base URL: https://agent-skill-hub.pages.dev. Full text: https://agent-skill-hub.pages.dev/llms-full.txt. JSON: https://agent-skill-hub.pages.dev/catalog.json. ## Skill - [ego lite (ego-browser skill)](https://agent-skill-hub.pages.dev/e/ego-lite/): The ego-browser agent skill and its open-source Node.js CDP harness for the ego lite browser: the agent runs ego-browser skill for a version-matched guide, then sends JavaScript that calls snapshot, fill, click, wait, navigate and capture inside its own isolated Space. - [live-panel](https://agent-skill-hub.pages.dev/e/live-panel/): Agent skill and stdlib Python renderer that turns one JSON config into an always-running terminal-style or pastel architecture diagram, with flowing packets, a scrolling log and counters, rendered to an H.264 mp4 or a self-contained looping web page. ## Skill pack - [Anthropic Agent Skills](https://agent-skill-hub.pages.dev/e/anthropic-skills/): Anthropic's public Agent Skills: document skills (docx, pdf, pptx, xlsx), skill-creator, mcp-builder, frontend-design, webapp-testing and more, plus the Agent Skills spec and a template. - [gstack](https://agent-skill-hub.pages.dev/e/gstack/): Garry Tan's skill pack that turns Claude Code and other agents into a virtual engineering team: office hours, CEO and eng plan reviews, design review, code review, QA with a real browser, security audit and ship. - [Superpowers](https://agent-skill-hub.pages.dev/e/superpowers/): A software development methodology delivered as composable skills: brainstorming, writing and executing plans, test-driven development, systematic debugging, subagent-driven development and verification before completion. ## Plugin - [Continual Learning (Cursor plugin)](https://agent-skill-hub.pages.dev/e/cursor-continual-learning/): Cursor plugin that keeps AGENTS.md up to date from transcript changes, using a stop hook plus a continual-learning skill that writes short high-signal bullets. - [pstack](https://agent-skill-hub.pages.dev/e/pstack/): poteto's Cursor plugin of rigorous engineering skills (poteto-mode, interrogate, tdd, create-verification-skill, no-comments and about 30 principle skills), also installable in Claude Code. - [Ralph Loop (Cursor plugin)](https://agent-skill-hub.pages.dev/e/cursor-ralph-loop/): Runs Cursor in a self-referential loop with afterAgentResponse and stop hooks, feeding the same prompt back until a completion promise appears or an iteration cap is hit. ## Plugin marketplace - [Claude Code Plugins Directory](https://agent-skill-hub.pages.dev/e/claude-plugins-official/): Anthropic's curated Claude Code plugin directory: internal plugins such as code-review, feature-dev, hookify, plugin-dev, LSP integrations and ralph-loop, plus partner plugins under external_plugins. - [Cursor Plugins](https://agent-skill-hub.pages.dev/e/cursor-plugins/): Cursor's official multi-plugin marketplace: pstack, thermos, ralph-loop, continual-learning, orchestrate, advisor, cursor-team-kit and integrations such as GitHub, Playwright and X. - [wshobson/agents](https://agent-skill-hub.pages.dev/e/wshobson-agents/): One catalog of 94 plugins with about 200 agents, 185 skills and 105 commands, adapted for Claude Code, Codex CLI, Cursor, OpenCode, GitHub Copilot and Pi. ## MCP server - [Chrome DevTools MCP](https://agent-skill-hub.pages.dev/e/chrome-devtools-mcp/): Lets a coding agent control and inspect a live Chrome through DevTools, as an MCP server or a CLI, for automation, debugging and performance traces. - [Cloudflare MCP Servers](https://agent-skill-hub.pages.dev/e/cloudflare-mcp-server/): Cloudflare's domain-specific remote MCP servers (docs, Workers bindings, observability and more) over Streamable HTTP; the README recommends the newer Cloudflare API MCP server. - [Context7](https://agent-skill-hub.pages.dev/e/context7/): MCP server and CLI that fetch current, version-specific library documentation and code examples into the agent's context. - [Firecrawl MCP Server](https://agent-skill-hub.pages.dev/e/firecrawl-mcp-server/): MCP server for Firecrawl: search, scrape and interact with the live web and get clean Markdown or structured data back. - [GitHub MCP Server](https://agent-skill-hub.pages.dev/e/github-mcp-server/): GitHub's official MCP server for reading repositories and code, managing issues and pull requests, and monitoring Actions runs. - [LinkedIn MCP Server](https://agent-skill-hub.pages.dev/e/linkedin-mcp-server/): Python MCP server that drives LinkedIn through the user's own browser session with Patchright: tools read profiles, companies, employees, jobs, the feed, posts and the inbox, and a smaller set sends messages, connection requests and job dismissals. - [MCP reference servers](https://agent-skill-hub.pages.dev/e/mcp-reference-servers/): Reference MCP server implementations maintained by the MCP steering group: everything, fetch, filesystem, git, memory, sequential thinking and time. - [Playwright MCP](https://agent-skill-hub.pages.dev/e/playwright-mcp/): MCP server that drives a browser with Playwright through structured accessibility snapshots, so models act on pages without screenshots. - [Serena](https://agent-skill-hub.pages.dev/e/serena/): MCP server that gives agents IDE-like semantic tools: symbol-level code retrieval, editing, refactoring and debugging through language servers. ## Agent harness - [Claude Code](https://agent-skill-hub.pages.dev/e/claude-code/): Anthropic's agentic coding tool for the terminal, IDE and GitHub. The public repository holds the changelog, issue tracker and example plugins. - [Codex CLI](https://agent-skill-hub.pages.dev/e/codex-cli/): OpenAI's open-source coding agent for the terminal, with IDE integrations; supports AGENTS.md and skills. - [OpenCode](https://agent-skill-hub.pages.dev/e/opencode/): Open-source terminal coding agent that works with many model providers, with a desktop app and an npm CLI. - [Pi](https://agent-skill-hub.pages.dev/e/pi-mono/): A minimal, extensible coding-agent harness: customise it with extensions, skills, prompt templates and themes; run it interactively, in print or JSON mode, over RPC or through its TypeScript SDK. ## CLI for agents - [agent-browser](https://agent-skill-hub.pages.dev/e/agent-browser/): Native Rust CLI that gives an agent a persistent browser: snapshot returns the page as compact text with element refs, and the agent acts on refs one command at a time. - [ccusage](https://agent-skill-hub.pages.dev/e/ccusage/): CLI that analyses coding-agent token usage and cost from local session data, with daily, monthly and session reports. - [lark-cli](https://agent-skill-hub.pages.dev/e/lark-cli/): Official Lark and Feishu CLI built for humans and agents, covering Messenger, Docs, Base, Sheets, Calendar, Mail, Tasks and more, with 24 bundled agent skills. - [MCP Inspector](https://agent-skill-hub.pages.dev/e/mcp-inspector/): Developer tool for inspecting MCP servers through a web UI, a scriptable CLI, a terminal UI and an experimental connection daemon. - [Playwright CLI](https://agent-skill-hub.pages.dev/e/playwright-cli/): Playwright command-line interface shipped with skills, aimed at coding agents that prefer concise CLI calls over large MCP tool schemas. - [Spec Kit](https://agent-skill-hub.pages.dev/e/spec-kit/): Toolkit that gives coding agents structured processes and templates, starting with spec-driven development, plus extensions for bug fixing and assessment. ## Agent SDK - [Claude Agent SDK for Python](https://agent-skill-hub.pages.dev/e/claude-agent-sdk-python/): Python SDK that exposes Claude Code's agent loop, tools, hooks and MCP support so you can build your own agents in Python 3.10+. ## Library - [Browser Use](https://agent-skill-hub.pages.dev/e/browser-use/): Open-source browser agent library that runs locally with any LLM; a paid cloud adds hosted browsers. - [HyperFrames](https://agent-skill-hub.pages.dev/e/hyperframes/): Open-source framework that renders HTML, CSS, media and seekable animations into deterministic MP4 video, with agent skills for Claude Code, Codex and Cursor. - [Stagehand](https://agent-skill-hub.pages.dev/e/stagehand/): TypeScript browser automation framework that mixes code with natural-language act, extract and observe steps, and runs a local browser or Browserbase. ## Model gateway - [Claude Code Router](https://agent-skill-hub.pages.dev/e/claude-code-router/): Local model gateway and control plane that gives Claude Code, Codex, OpenCode, Pi and other clients one endpoint, with routing, failover and observability across providers. - [magpie](https://agent-skill-hub.pages.dev/e/magpie/): Menu-bar app and local gateway that serves every coding agent's model from one place and moves to the next account when a quota runs out. ## Curated list - [Awesome Claude Code](https://agent-skill-hub.pages.dev/e/awesome-claude-code/): A hand-picked list of Claude Code resources: skills, hooks, slash commands, status lines, orchestrators and applications, with a recently-added section. - [Awesome Claude Code Subagents](https://agent-skill-hub.pages.dev/e/awesome-claude-code-subagents/): A collection of Claude Code subagent definitions for specific development tasks, installable as a Claude Code plugin. - [Awesome Claude Skills (Composio)](https://agent-skill-hub.pages.dev/e/awesome-claude-skills-composio/): A large list of Claude Skills and plugins, with some skills stored in the repo (skill-creator, mcp-builder, file-organizer, content-research-writer and others). - [Awesome Claude Skills (travisvn)](https://agent-skill-hub.pages.dev/e/awesome-claude-skills-travisvn/): A curated list of Claude Skills, resources and tools, with an explanation of how skills load through progressive disclosure. - [Awesome MCP Servers](https://agent-skill-hub.pages.dev/e/awesome-mcp-servers/): A curated list of Model Context Protocol servers grouped by category, plus clients and other MCP resources. - [Claude Code Templates](https://agent-skill-hub.pages.dev/e/claude-code-templates/): An open catalog (aitmpl.com) of about 1,900 Claude Code components: skills, agents, slash commands, MCP integrations, settings, hooks and project templates, installable with one command. - [MCP Registry](https://agent-skill-hub.pages.dev/e/mcp-registry/): The official registry service that gives MCP clients a list of MCP servers, with a public API at registry.modelcontextprotocol.io. ## Spec or format - [AGENTS.md](https://agent-skill-hub.pages.dev/e/agents-md/): A simple open format, a README for agents, that gives coding agents a predictable place for project context and instructions.