xingkongliang/skills-manager

Skills Manager – Quick Overview

What it is – A desktop (macOS, Windows, Linux) and CLI tool that lets you collect, organise, and deploy AI‑agent “skills” (code snippets, prompts, extensions) across many coding assistants such as Claude Code, Cursor, GitHub Copilot, Gemini CLI, etc.

Why it matters – Modern AI coding agents rely on small, reusable skill packages. Keeping those packages in sync, version‑controlled, and correctly placed for each agent quickly becomes a pain point. Skills Manager provides a single library, a visual dashboard, and a Git‑backed backup that solves the "where does this skill live?" problem for both humans and the agents themselves.

---

Core Features (as described in the README)

Feature	What you can do	How it helps
Unified skill library	Store skills from Git repos, local folders, .zip/.skill archives, or the skills.sh marketplace in one central directory (~/.skills-manager by default).	No more hunting for scattered skill folders on different machines.
Marketplace	Browse and search popular skills directly in the app.	Discover ready‑made extensions without leaving the UI.
Agent‑aware management	Agents (Claude Code, Codex, Cursor, …) can install, deploy, or query skills through the manager instead of editing their own folders.	Guarantees metadata, presets, and update tracking stay consistent.
Presets	Group skills into named sets; activate/deactivate a preset with one click.	Quickly switch a whole toolbox for a project or a specific agent.
Global & Project workspaces	View all skills an agent sees globally, or manage project‑local skill folders and sync them with the central library.	Keeps global defaults and per‑project overrides in sync.
Linked workspaces	Point the manager at any arbitrary directory to treat it as a skill root.	Works with tools that store skills outside the default paths.
Multi‑tool sync	Sync a skill to any supported agent via symlink or copy from the card view.	One‑click deployment, visual status badges.
Batch operations & tagging	Multi‑select for bulk enable/disable, export, delete; tag skills and filter by tag/source.	Saves time when handling dozens of skills.
Update tracking	Detect upstream changes for Git‑based skills, re‑import local ones.	Keeps skills up‑to‑date automatically.
Backup & multi‑device sync	Connect a private GitHub repo (or any Git remote) to version‑control the whole library; automatic merges are skill‑aware, conflicts never block.	Seamless work across laptops, desktops, or CI machines.
Activity log & export	Export logs for troubleshooting.	Easy bug reports.
In‑app updates	Notifies of new releases; installs on macOS & Windows with a click.	Stay current without manual downloads.
Custom tools	Add agents/tools with custom skill directories.	Extensible beyond the 54 built‑in agents.

---

How to get started

1. Install – brew install --cask skills-manager on macOS or download the appropriate installer from the GitHub releases page for Windows/Linux.
2. Open the app – The default library lives at ~/.skills-manager. Use the Global Workspace to pick an agent (e.g., Claude Code).
3. Add skills – Click + Add Skills, browse your library or the marketplace, tick the agents you want, and hit Add.
4. Use presets – Create a preset (e.g., React‑Best‑Practices) and enable it for a workspace with a single pill click.
5. Sync across devices – In the Backup sidebar, sign‑in with GitHub to create a private repo; the app will now auto‑commit and push changes.
6. CLI – The same core engine is exposed as skills-manager-cli. Example: skills install ./my-skill && skills deploy my-skill --agent claude_code.

---

Technical snapshot

• Frontend: React 19 + TypeScript, built with Vite, styled with Tailwind CSS.
• Desktop shell: Tauri 2 (Rust‑based, lightweight native wrapper).
• Backend core: Rust, persisting data in SQLite (rusqlite).
• Internationalisation: react-i18next.
• CLI: Rust binary bundled with the app; also published separately for direct shell use.
• License: MIT.

---

Who might benefit?

• Developers who use multiple AI coding assistants and want a single source of truth for their skill packages.
• Teams that need to share a curated set of skills across machines and keep them version‑controlled.
• Tool builders who want their agents to be able to install/manage skills programmatically.

---

All details above are taken directly from the repository’s README; no additional features have been inferred.