skillsync-mcp / site

# SkillSync MCP

> Security-gated skill management for AI coding assistants. The only MCP server that gates skill installation behind a full security scan.

## Overview

SkillSync MCP (`@stranzwersweb2/skillsync-mcp`) is an open-source Model Context Protocol (MCP) server that provides security-gated skill management for Claude Code and 6 other MCP-compatible clients. It scans GitHub skill repositories for 60+ threat patterns across 15 categories before allowing installation, permanently blocking critical threats like prompt injection, reverse shells, credential theft, and supply chain attacks.

- **Version**: 1.3.0
- **License**: MIT
- **Author**: Aditya Sugandhi
- **npm**: https://www.npmjs.com/package/@stranzwersweb2/skillsync-mcp
- **GitHub**: https://github.com/adityasugandhi/skillsync-mcp
- **Website**: https://skillsync.js.org/

## Install

```bash
npx -y @stranzwersweb2/skillsync-mcp
```

Or add to Claude Code's `~/.claude/settings.json`:

```json
{
  "mcpServers": {
    "skillsmp": {
      "command": "npx",
      "args": ["-y", "@stranzwersweb2/skillsync-mcp"]
    }
  }
}
```

## 13 MCP Tools

### Search & Discovery
- `skillsmp_search` — Keyword search across SkillsMP marketplace
- `skillsmp_ai_search` — AI-powered semantic search via Cloudflare AI
- `skillsmp_search_safe` — Search + auto-scan top results for security threats
- `skillsmp_suggest` — AI-powered skill recommendations based on installed skills

### Security
- `skillsmp_scan_skill` — Security scan a GitHub skill repo (60+ patterns, 15 categories)
- `skillsmp_compare` — Side-by-side security comparison of two skills
- `skillsmp_audit_installed` — Deep security audit of a specific installed skill

### Skill Management
- `skillsmp_install_skill` — Scan then install to ~/.claude/skills/ (blocks critical threats)
- `skillsmp_uninstall_skill` — Remove an installed skill
- `skillsmp_list_installed` — List all installed skills with risk levels

### Cloud Sync
- `skillsync_configure` — Manage sync subscriptions and settings
- `skillsync_sync_now` — Run sync cycle (install/update/remove skills)
- `skillsync_status` — Show sync engine status

## Supported Clients (7)

Claude Code, OpenClaw, Cursor, Windsurf, GitHub Copilot, Zed, nanobot

## Security Model

Every installation is gated by a multi-level security scan:

- **Safe/Low risk**: Install proceeds with warnings
- **Medium/High risk**: Blocked, requires force=true
- **Critical risk**: Permanently blocked, no override

### 15 Threat Categories Detected

Prompt Injection, Unicode Homoglyphs, Destructive Shell, Remote Code Execution, Reverse Shells, Credential Theft, Supply Chain, Privilege Escalation, Exfiltration, Obfuscation, Code Execution, Shell Execution, Network, Crypto Mining, Dotfile Poisoning

### Safety Guards

- SSRF prevention (only github.com URLs)
- Path traversal prevention
- npm install --ignore-scripts
- Max 50 files, 512KB per file, 2MB total
- SHA-256 content hashing for TOCTOU verification
- Output sanitization (strips zero-width Unicode, bidi overrides)
- 15-minute scan cache with LRU eviction
- fs.watch startup verification

## Tech Stack

TypeScript, Node.js 20+, @modelcontextprotocol/sdk, Zod validation, MCP over stdio