Agent Skills
Catalog of official agent skill packages, collected from confirmed official URLs.
Forge: A Model-Agnostic AI Coding Harness and CLI — A Decision Guide Before Your First Use
Forge is best suited for model-agnostic AI coding tasks in the terminal, automatically routing each task to the optimal model by cost and capability. Getting started requires accepting the one-line shell install script (which also installs the desktop app by default — decline with n) and completing forge setup on first launch to connect models and security settings.
Native Rust binary (no Node/Python/Bun runtime or Electron)terminal UI (full-screen or inline)one-shot/headless CLInative desktop appiOS/Android/Web companion appOpenAI-compatible local APIAI Maestro: A Safe Getting-Started Guide to the Multi-Agent Orchestration Dashboard
For developers who want to manage multiple AI coding agent sessions (Claude Code, Codex, Grok Build, etc.) from a single web dashboard. The most important prerequisite before use: it has no application-layer authentication, so it must only run on a trusted network, and port 23000 must not be exposed to untrusted networks.
Node.js Web 仪表盘，通过 tmux 管理代理会话（xterm.js + node-pty + WebSocket）Claude CodeCodexGrok BuildREA: Reverse Engineer Native Binaries, Electron Apps, .NET Assemblies, and Websites with an Agent
Best suited for letting an agent reverse engineer native binaries, JavaScript/Electron apps, .NET assemblies, and websites via an MCP server. Before use, you must confirm the Node.js version requirements (22.x >=22.23.4, 24.x >=24.22.1, or 26.12.1+), and note that setup rewrites agent configuration while native analysis depends on an existing Hopper/Ghidra/IDA.
Node.js (MCP server)VibesPlug Pro: A 176-Skill Multi-Agent Development Suite
Install VibesPlug Pro only if you want to enable 176 skills and an autonomous multi-agent development workflow uniformly across multiple IDEs — and only if you accept that the official installation method executes a remote shell script from the main branch, and are willing to review the script first or run it in an isolated environment.
Shell installer on local machine; runs as IDE plugin and MCP v1.x server (stdio & SSE)AntigravityClaude CodeCursorWindsurfOpenAI CodexSkills-Atlas: Browse public Agent Skills by topic and install them in bulk
For developers who want to select and install several public Agent Skills in one go, provided the machine runs Node ≥ 20 and, for each skill, you confirm before installing what it will execute, what it will read, and what it will send outward.
Node.js ≥ 20claude-codeTrending Claude Skills: A Leaderboard of claude-skills and AI Agent Repositories Refreshed Every 15 Minutes
For anyone who wants to quickly discover recently active claude-skills, MCP server, and AI agent capability repositories: the leaderboard is fully rewritten on a schedule by GitHub Actions from GitHub Search API results, so reading the table in the repository's README is all it takes, and nothing is written to your local disk.
GitHub Actions 定时任务，每 15 分钟刷新一次榜单Natively: Real-Time Screen Overlay AI for Interviews and Meetings on macOS
Suited to live Q&A and system design drawing through a screen overlay during interviews, sales pitches, and on-site meetings — macOS desktop only, with the free tier limited to personal and non-commercial use.
macOS 桌面客户端（界面按 macOS 原生风格设计）Shikhu: Turn Your Codebase into Self-Tests and Track Your Knowledge Coverage
Shikhu is for developers who want AI coding agents to drill them on their own code and to track knowledge coverage over the long term. The prerequisite is bringing your own OpenRouter API key; without it, no step that involves model generation can run.
命令行工具（uv tool 隔离安装，独立于项目依赖）Claude CodeCodexCursorOpenCodeOpenDesigner: ship a design system for your project with a few answers
For teams that need a consistent visual specification for an AI coding project in progress: after installing the plugin, answer at most five short questions and you get DESIGN.md, PRODUCT.md, and an opendesigner/ token output. The prerequisite is Python 3.10 or later on your machine.
Python 3.10 或更高版本；引擎为单个 Python 文件，无需安装依赖Getting Started with Gentle-AI: Adding Memory, Workflow, and Evidence to an Existing AI Coding Agent
For developers who already use an AI coding agent and want it to remember project conventions over the long run and leave verifiable records, provided you first have a supported agent to host the configuration.
已有 AI 编码代理之上的配置层（不自带代理或 IDE）Claude CodeCursorOpenCodeCodexPiPaperclip: The Open-Source Orchestration Console That Runs a Bunch of AI Agents Like a Company
Suited to teams that want to pull self-supplied agents such as Codex, Claude, Cursor, and OpenClaw under one shared goal, dispatch work uniformly, watch progress, and control cost — provided you bring your own agent accounts and model quota, and run the Node.js service locally following the Quickstart.
Node.js 服务端 + React 界面OpenClawCodexClaudeCursorcodeArbiter Decision and First-Use Guide: Bringing Gated Code Governance to Claude Code, Codex CLI, and Pi
Best suited for adding a layer of gated code governance to repositories built with Claude Code, Codex CLI, or Pi: every intent goes through a gated skill or reviewer agent, nothing commits until all gates are green, decisions run through SMARTS, and the audit trail is append-only. The plugin stays dormant after install and only starts working after you explicitly opt a repository in.
Claude Code / Codex CLI / Pi plugin suite; requires Pi 0.80.10, Node.js 22.19 or newer, Python 3, and Git; supports Windows, macOS, and LinuxClaude CodeCodex CLIPicareer-ops: the job search assistant that runs locally in your AI coding CLI, drafting applications only — pressing Submit is always up to you
career-ops suits job seekers who want to evaluate openings and track applications locally inside their own AI coding CLI (such as Claude Code or Codex); it only drafts and fills, never submits automatically, and whether to apply is always your call.
Local AI coding CLI (Claude Code, Codex, OpenCode, Gemini / Antigravity, Grok, Qwen and more) via the open Agent Skill StandardClaude CodeCodexGemini / AntigravityOpenCodeGrokQwenGentle Shell: judging whether this Pi coding-agent harness suits you, and making a first attempt
gentle-shell best suits developers who already use the Pi coding agent and want the agent to explore before changing anything while small tasks stay in small, controlled steps. Before adopting it, two things must be accepted: installing through the npm package gentle-pi replaces Pi's builtin codemode extension with a compact renderer, and both the package and the repository remain named gentle-pi until the migration completes.
Pi coding-agent sessionPiBrigade: Sync tools and memory across harnesses, with one local-file record of your coding agents' work
Best suited to users who need to keep tools and memory in sync across different coding agents (harnesses) and share one local work record; before use you need Node.js 22+ and npm, and you should be aware of the officially disclosed passive version request telemetry side effect.
Node.js 22 或更新版本 + npmDuology: Claude Code skills for iPhone Duo foldable app development
Best for developers who already meet the stated requirements (a Mac with Apple silicon, Xcode 27.1, Claude Code) and want to design and check apps for the iPhone Duo foldable: with the `duology` plugin installed, an agent can run `duo-lint` and the 15-point checklist and return the issues to fix ordered by severity; Cursor, Codex, Copilot, Gemini CLI and similar clients install skills only, and slash commands plus the lint hook remain Claude Code features.
Claude Code；需 Apple silicon 的 Mac 与 Xcode 27.1 (27A9275 or later)Claude CodeCursorCodexCopilotGemini CLIMSP Skills: Query PSA, RMM, Backup, and M365 Data in Claude Code with Natural Language
Best suited for MSP engineers who want to query PSA, RMM, backup, and M365 data in Claude Code using natural language: every connected system syncs into a local SQLite mirror with full-text search, so queries are instant and work offline; before starting, you must prepare your own credentials for the systems involved, which are supplied at runtime and not stored in the repo.
Claude Code / Codex（Skill）；Claude Desktop / ChatGPT / Cursor 等（MCP server）ClaudeChatGPTCodexCursorWindsurfClaude CodeClaude DesktopOpenRig: Turn Claude Code and Codex into a YAML-Defined Agent Team
Choose OpenRig when you want to organize Claude Code and Codex into an agent team defined in YAML and booted with a single command; have Node.js 22 or 24 and tmux ready before you start, and note that rig setup appends to ~/.tmux.conf and installs Herdr by default (--no-herdr skips it).
Node.js 22 或 24 + tmux（macOS / Linux / Windows WSL2）Claude CodeCodexGAIA: A Spec-First Development Harness for Claude Code
Best for developers who use Claude Code for serious software development and want quality constraints on every commit and merge: GAIA enforces specs before code, sets a quality gate on every commit, and brings a code audit into every merge. It requires Node.js >= 22.22.0, and officially projects are created with an npx command rather than by cloning or forking.
Claude Code (Node.js >= 22.22.0)Claude CodeSkillSpector: Security Scanner for Agent Skills
Best suited for a static security pre-check before installing an agent skill: it detects vulnerabilities, malicious patterns, and security risks. Prerequisites are uv on your machine, and for MCP-based invocation a local agent that supports MCP.
Python (uv tool)Claude CodeCodex CLIGemini CLIgccli: A Command-Line and Agent Query Guide for Garmin Connect Data
gccli is best for querying your Garmin Connect activities, health, and body composition data directly from the terminal or an agent client, and handing parseable JSON or plain/TSV output to scripts. Garmin SSO authentication is required before use, and the fallback token file is encrypted with a fixed passphrase embedded in the source code — make sure you can accept that before starting.
CLI；从源码构建要求 Go 1.26 或更高Claude CodeCodexHermesOpenClawCotal: an open standard that lets multiple AI agents share one skill set and resume workflows across machines
Best suited for people who want to share one set of Agent Skills across multiple agent clients, or who need to resume an unfinished workflow on another machine; before getting started you need to accept an official install script that writes into your home directory, and key facts such as system requirements and how to uninstall are not documented on the official page.
终端 CLI（cotal 命令行），配合受支持的 agent 客户端使用CodexCursorOpenCodeGemini CLIWindsurfClaude CodeCurie: The open-source, self-hostable platform that turns coding agents into co-workers
Best for development teams that want to delegate GitHub issues to an AI coding co-worker and get back a PR/MR awaiting review; you need to bring your own Docker/Kubernetes environment and an Anthropic or OpenRouter model credential.
Docker + kind (Kubernetes) + kubectl + HelmBinance Skills Hub: Crypto Skills Marketplace for AI Agents
Best for letting AI agents search tokens, execute trades, track wallets, monitor signals, and interact with DeFi protocols through natural language. Requires Node.js 22 or higher, and some endpoints also require Binance API credentials.
Node.js 22+OpenClawClaude CodeOmO (omo command): turn tasks into a parallel agent graph with mass ulw
OmO suits handing deep research, decks, or frontend/backend engineering work end to end to a single `omo` command that runs in parallel; getting started requires a subscription to one of Claude, ChatGPT, Kimi, or GLM with sign-in via /login, and note that the npm package named omo is unrelated.
CLI，运行于 senpi（pi 的 fork）ClaudeChatGPTKimiGLMIDD Skill: An Issue-Driven Development Workflow Template
Best for letting AI coding agents run the full development loop in your own Git repository, from claiming an issue to merging a PR; before use, prepare git, an authenticated gh CLI, jq, and curl, and confirm the default fully autonomous merge policy fits your team's norms.
AI 编程 agent 会话GitHub CopilotClaude CodeOpenAI Codex CLIOpenCodeAntigravity CLIconsole-skill: 11 Claude skills for reverse-engineering your own PS2 disc and porting it to PC
For people who have imaged their own PS2 disc and want to use Claude to reverse-engineer that disc and recompile it to run on PC — provided you image the disc yourself and build the PS2Recomp runtime yourself.
Claude / Claude CodeClaudeClaude Codenaskah-id: An AI Agent Skill for Indonesian Academic Writing and Editing
Best suited for writing and editing Indonesian academic texts such as journal articles, theses, and grant proposals in line with EYD V and KBBI standards; requires an agent environment with Node.js 18+, and it strictly never fabricates data or references, leaving a to-be-filled marker where citations are missing.
Node.js 18+ (via npx skills); optional local scripts need Python 3.7+Claude CodeCodexCursorGemini CLIGitHub CopilotWindsurfvideo-to-blender: Turn a reference video, image, or idea into an editable procedural Blender node system
This coding agent Skill is best when you want to generate an editable procedural Blender node system (geometry, shading, compositor nodes) from a reference video, image, or idea; before starting, you must have Blender 5.2+ ready and confirm the agent is installed, otherwise its build-and-validate workflow cannot be driven.
Blender 5.2+ driven by coding agents (Claude Code, Codex, Cursor, Gemini CLI, Copilot)Claude CodeCodexCursorGemini CLICopilotCodex Security: Traceable Code Vulnerability Scanning in Your Terminal
Suited to running vulnerability scans against a specific directory or a single Git change from the terminal and producing source-level evidence, provided the global npm install and login are done; in CI scenarios an API Key or device-authorized login is also required.
Node.js 22.13.0+ within 22.x, or Node.js 24.x or 26.x; Python 3.10+ (Python 3.10 also requires tomli)CodexGoing Dark: A Strict Method for Getting AI Coding Agents from Issue to Release
If you want AI coding agents to run your software project through one strict pipeline — issues written as specs, handoff prompts for any agent, audits that fix what they find — Going Dark is for you. First confirm whether your agent is Claude Code (marketplace path) or another agent that reads AGENTS.md (npx installer); the two install paths are completely different.
Markdown 技能（Agent Skills 格式）；npx 安装路径需 Node.js 20+Claude CodeOpenAI CodexGemini CLIGitHub CopilotCursor任何读取 AGENTS.md 的 agentpapercut-explainer: batch-produce vertical papercut-style explainer videos in Claude Code
papercut-explainer works best as a staged pipeline in Claude Code, from strategy to QA, producing vertical papercut-style explainer short videos. The catch: you must have Node 20+, ffmpeg, rsync, and Python 3 set up, and full features require cloning the repo and running install.sh — not just installing the skills.
Node 20+ / Python 3（Claude Code 前端）Claude CodeFactory: Turn Your Factory into a Runnable 3D Discrete-Event Simulation with One Sentence
Factory is best at turning a one-sentence factory description into a runnable 3D discrete-event simulation page for quickly validating line designs; before use, confirm your machine has Node 20+, a WebGL2-capable browser, and Claude Code installed.
Node.js 20+ (scripts), browser with WebGL2 (output page)claude-codePentest Projects Discovery: Generate Hands-On Penetration Testing Project Roadmaps with Claude
Best for anyone who wants to learn penetration testing systematically through hands-on projects in a legally authorized environment: create a Claude Project on Claude.ai, add the repository's SKILL.md, then request roadmaps; it is intended for authorized learning purposes only.
Claude.ai (Claude Projects / Skills)Claude.ai (Claude Projects)Awesome Reverse Engineering: Pick Reverse Engineering Tools and Learning Resources by Task
Best for quickly finding reverse engineering tools and learning resources by task, especially for wiring Ghidra/radare2-style analysis into an AI agent via MCP; it is only a curated list with no install commands, so real tools and prerequisites must be confirmed in each upstream project's docs.
无运行时（精选文档清单，非可安装工具）ship-ready: Ship AI-written Code with Evidence
ship-ready suits developers using Claude Code, Codex, Copilot, or Gemini CLI who want AI-written code turned into an evidence-backed "ready to ship" state—provided you first run the official install script and accept it writing a guardrails marked block into the agent instruction files.
bash/git installer; agent skill files for Claude Code, Codex, Copilot, Gemini CLIClaude CodeOpenAI CodexGitHub Copilot (CLI and VS Code)Gemini CLI / AntigravityLuobo PPT: Generate and Beautify Editable Native PPTX in Your Agent
For turning an article or topic into an editable native PPTX — or beautifying an existing .pptx — inside Agents that can execute commands, such as Claude Code, Codex, Gemini CLI, and Cursor. Before use, confirm local dependencies like Python 3.8+ and a Chinese font; the first run automatically installs missing Python packages.
Python 3.8+（Agent Skill，SKILL.md）Claude CodeCodexGemini CLICursorcinetic: Let Coding Agents Direct Cinematic Launch Short Films
cinetic is best when you want your coding agent to turn a one- or two-sentence brief directly into a cinematic launch short film or motion design — but only if your machine first has Node 22+, ffmpeg 6+, Python 3.11+, and headless Chromium in place, otherwise the rendering pipeline cannot run.
Node 22+, ffmpeg 6+, Python 3.11+, headless Chromium; macOS or Linuxawesome-claude-video-skills: A Safety-Rated Checklist for Picking Video Skills for Coding Agents
This awesome list is best used before installation to help you pick from 168 open-source Skills that let coding agents like Claude Code and Codex create videos, filtering by video type and safety rating — but it is not a runnable tool, has no package installer, and you must verify each collected repository's README safety rating before installing it.
无运行时（静态 awesome 清单，消费方式为阅读 README、在线筛选页与 data/skills.json）logo-design-skill: Turn Claude into a Disciplined Logo Designer
Best suited for logo design work with Claude, from design brief to production-ready SVG files and brand guidelines. Before use, confirm you're on Claude Code with Python 3.8+ and, optionally, an SVG renderer.
Claude（Claude Code 插件；Claude.ai / Claude Desktop 需上传 Skill 压缩包）+ Python 3.8+Claude CodeClaude.aiClaude Desktopopus-js-animations: Direct and render frame-accurate JavaScript films with Opus inside Claude Code
For using Claude Opus 5.5 in Claude Code to develop a one-line idea into a frame-accurate MP4 animated short; prerequisites are Node ≥ 22, headless Chrome, ffmpeg, and Python 3 + numpy, and nothing is built until you say go.
Claude Code (Claude Opus 5.5) + Node ≥ 22, headless Chrome/Chromium, ffmpeg, Python 3 + numpyClaude Codeoctop-browser: Make Your Agent's Browser Operations More Accurate and Reliable
octop-browser is best for letting MCP-capable agents complete browser automation more accurately and reliably through real Chromium. Before use, you must accept that it downloads a Chromium binary (possibly one-time bootstrapping Playwright) and adopt the dom-tree→ref safe locating pattern.
Python 3.11+，MIT 许可；直接 CDP 驱动真实 Chromium，不依赖 Playwright 运行时MCP-capable agentsReverse-Engineer a Server Protocol from a Game Client and Reproduce a Deployable Server
For reverse-engineering work that starts with only a client and aims to reconstruct the protocol and run the original client locally offline; establish an evidence inventory and protocol.spec.yaml first, then consider deriving the server.
Agent SkillClaude CodeCursorCodexGitHub CopilotWindsurfGeminiClineswift-testing-pro: Helping Coding Assistants Write Better Tests with Swift Testing
In an environment with Node already installed, use the official npx command to install swift-testing-pro so Claude Code, Codex, Cursor, or Gemini can help write Swift Testing tests; review the suggestions before allowing file changes.
Swift TestingClaude CodeCodexCursorGeminiOnchainOS Skills: Check tokens and balances first; don't swap right away
A good fit for people who want a coding assistant to look up tokens, market data, or wallet balances, but every skill requires OKX API credentials and swaps or transaction broadcasts really use funds.
npxClaude CodeCursorCodex CLIOpenCodeFlue: install with pip, then run setup to confirm your assistant sees the bridge
For people who want assistants such as Codex or Claude Code to drive desktop software like Photoshop and Blender from the terminal: run pip install flue and then flue setup, and only confirm that the assistant knows Flue is available on the first try—do not send a script to the app yet.
pipCodexClaude CodeGemini CLIAntigravity CLIOpenCodeTurn technical documentation into clear English with simple-english
For people who want an agent to write clear technical English under ASD-STE100 discipline: install it first with the official npx command, then make a first rewrite with rewrite this with simple-english.
npx skillsClaude CodeCursorVS Code CopilotOpenAI CodexGemini CLIGooseOpenCode01coder Agent Skills: Confirm codex first, then make one verifiable delegation
For people who want to use VerySmallWoods skills in an assistant that supports skills.sh; run only the official npx installation, confirm that codex is installed, then give a clearly scoped task to codex exec and check the result.
npxClaude CodeCodexCursorWindsurfAwesome LLM Apps: give your agent project-graveyard in ten seconds
Install project-graveyard inside Claude Code, Codex, or Cursor with the official example command, then check that the skill took effect by asking that one English question; do not invent other install URLs from the collection's directory tree.
coding agent skillsClaude CodeCodexCursorHono Skill: Install from the relocated honojs/skills repository
For anyone who still wants to install the Hono skill: this repository has moved to honojs/skills; use only the official npx command, and remove this repository first if you previously added it as a marketplace.
npx