Transcript
[00:00] Here’s a question for you and don’t panic, but how many AI skills do you have installed on your machine right now? More importantly, how many of them are you actually using? It turns out it’s a really hard question to answer because you’re going to have to go dig through all of your agent directories, all of your project directories, any custom directories where you’ve got your own custom skills, any remote servers where you’re storing skills. SkillsGate attempts to solve this problem with a very simple unified interface that tells you which skills are on your machine, which agents and projects are registered with. It gives you a way to discover new skills and install them, even to edit those skills if you want, and to create new skills.
[00:46] Before we jump into SkillsGate, this is the part where I tell you to subscribe. If you like this content, subscribe. So SkillsGate is from a gentleman by the name of Sultan Valiyev. Here is the SkillsGate repository. If we scroll down here, it says it’s a visual skill manager for AI agents. It’s got a desktop app and a TUI. We are going to be focused on the desktop app. The TUI is great. It works. I just wasn’t in love with it. But if you want to install it, you can NPX SkillsGate. And just to be clear, I mean, it’s very well done. I think it’s got parity with the desktop app. I think the reason I didn’t like it is because it took me forever to figure out that Control Q was the way to quit it. So I might revisit that.
[01:27] We’re going to jump over to SkillsGate.ai, where you can see it claims 91,000 public skills because it’s pulling them all from SkillsSH, which is the same resource behind NPX Skills. We can download the desktop app, which is going to lead us back to the repository and the releases page. And you just want to make sure you download the correct release for your system. So in my case, that’s going to be this ARM64 DMG for an M3 MacBook. When we load up the application, we’re dropped into the Skills library. We’re going to circle right back to this. I’m going to start off with settings where we can update the default scope, either global or per project. We can choose if we want to copy the Skills into the agent directories or if we want to symlink them. There’s some search preferences here I haven’t really figured out. I turned off telemetry because that’s just my prerogative. We can check for updates. We can also add. So if you’ve got skills in a custom directory, you can add a directory here and it’ll scan that as part of its process of displaying your skill library. There’s some additional stuff here. You can take your time and look through all those.
[02:39] So when we get to the installed skills, you’ve got your different agents here and you can see which agents are actually installed for. More importantly, we can come over here to discover and we can search for new skills. So for example, I’ve been wanting to take a look at Vercel’s native SDK. So I can search here and there it is. I can click on that guy and I can choose to install it. I tend to stick with the universal agent. So I’m going to click on install there and it’s basically just running that MPX skill ad for me. So now that that’s installed, I can come over here and I can see that it’s there along with this native UI and zig skill that came with it. I can browse the skill. I can actually go in and edit the skill if I want to. And we’ll take a look at that in an agent in just a second. We can also jump over here to servers and if you have a server, a local server on your network where you’re keeping skills, you can go ahead and add that here, throw on a username or an SSH key path, and you will have access to all the skills on your network.
[03:49] So I created a directory here called native test. I’m going to open up open code in that directory. I’m going to check my skills and we can say the native SDK skill is there. Let’s say scaffold a new native project here. We’ll see if that works. Cool. So it’s obviously reading the skill. And this is one thing I really enjoy doing when I want to learn about something and there’s a skill there. I can say, set it up for me so that I can get in and take a look at it. So this is a brand new project. I think it’s a month or two old. There are a couple rough edges. It’s not necessarily picking up on all of my agents. Again, I don’t really care that much about that because I tend to use the universal like dot agents directory for all my skills, but it’s not picking up on like Hermes and some other guys that I’m using, but it is a electron app. So we get kind of all the stuff you get in a browser. I can zoom out and zoom in. I can bring up the dev tools if I want. So it comes with a few caveats, but overall, I feel like this is, I don’t know why nobody else has done this before. It’s such a simple idea that all of a sudden makes managing the skills on my machine a lot easier to do.
Description
SkillsGate is a visual skill manager for AI coding agents. It lets you browse, install, and manage reusable AI agent skills (SKILL.md files) from a single interface. It integrates with skills.sh to discover 91,000+ public skills, and supports 18+ agents including Claude Code, Cursor, Windsurf, GitHub Copilot, Codex CLI, Cline, Continue, Roo Code, Zed, Aider, and more. Available as both a desktop app (Electron) and a terminal UI (TUI).
Details
| URL | https://github.com/skillsgate/skillsgate |
| Type | Desktop / CLI (TUI) / Web |
| Pricing | Free |
| Open Source | Yes |
| License | MIT |
| Tech Stack | TypeScript, Electron, React Router v7, Cloudflare Workers, SQLite (better-sqlite3), Bun, CodeMirror |
| Platforms | macOS (Apple Silicon + Intel), Windows, Linux |
| Self-Hosted | No (skills.sh index is external; local DB is local) |
Key Features
- skills.sh integration — browse and search 91,000+ public skills directly from the app
- Per-agent management — install or remove skills on a per-agent basis (e.g., install to Claude Code but not Cursor)
- Built-in editor — view rendered skill content or edit raw SKILL.md source with a CodeMirror editor, saved to disk instantly
- Remote servers — connect to machines via SSH to browse, install, and sync skills across a fleet
- Private skills — keep skills local to your machine or share them with your team
- Favorites — star skills from the catalog for quick access
- Settings sync — desktop and TUI share preferences via a local SQLite database
Best For:
Developers and teams managing multiple AI coding agents who want a unified way to discover, install, and manage agent skills (SKILL.md / .cursorrules-style instructions) without manually hunting through GitHub repos.
Integrations
skills.sh (public skill catalog), SSH (remote server management), 18+ AI agents (Claude Code, Cursor, Windsurf, GitHub Copilot, Codex CLI, Cline, Continue, Amp, Goose, Roo Code, Zed, Aider, OpenCode, Kilo Code, Trae, Pear AI, OpenClaw, Universal), Cloudflare Workers (web app hosting).
Notes
- Monorepo managed with npm workspaces: apps/web (React Router v7 on Cloudflare Workers), apps/desktop (Electron), packages/cli (Node CLI, npm skillsgate), packages/tui (Bun TUI, @skillsgate/tui), packages/ui (shared React components), packages/local-db (SQLite persistence + SSH client).
- Requires Node.js 18+ for development; Bun is needed for TUI development.
- Desktop app uses native module better-sqlite3 — may need manual rebuild (npm run rebuild:native —workspace=@skillsgate/desktop) if scripts are skipped.
- The TUI can be run via npx skillsgate with no install, or npm install -g skillsgate globally.
- Built by Sultan Valiyev.