SkillSync MCP
The only MCP server that gates skill installation behind a full security scan.
SkillSync MCP is a security-first skill management server for AI coding assistants. Before any skill is installed, it runs a comprehensive threat analysis across 60+ patterns covering 15 attack categories.
Key Features
Security Scanning
Every installation is gated by a multi-level security scan across 60+ threat patterns in 15 categories: prompt injection, Unicode homoglyphs, reverse shells, credential theft, supply chain attacks, crypto mining, RCE, obfuscation, and privilege escalation. Critical threats are permanently blocked with no override possible.
Cloud Sync
Subscribe to search queries and auto-sync skills across machines. Configurable risk thresholds, conflict policies, and dry-run previews. Lock file support for managed vs manual skill tracking.
AI-Powered Search
Keyword and semantic search across the SkillsMP marketplace powered by Cloudflare AI. Smart suggestions based on your installed skills. Side-by-side skill comparison with security diffs.
Skill Management
Install, uninstall, list, and audit skills. Startup verification with fs.watch and SHA-256 content hashing. TOCTOU protection. npm install --ignore-scripts for supply chain safety.
13 MCP Tools
| Tool | Category | Description |
|---|---|---|
skillsmp_search | Search | Keyword search |
skillsmp_ai_search | Search | AI semantic search |
skillsmp_search_safe | Search | Search + auto-scan |
skillsmp_suggest | Search | AI suggestions |
skillsmp_scan_skill | Security | Security scan |
skillsmp_compare | Security | Side-by-side compare |
skillsmp_audit_installed | Security | Deep audit |
skillsmp_install_skill | Management | Install skill |
skillsmp_uninstall_skill | Management | Uninstall skill |
skillsmp_list_installed | Management | List installed skills |
skillsync_configure | Cloud Sync | Configure sync |
skillsync_sync_now | Cloud Sync | Run sync |
skillsync_status | Cloud Sync | Sync status |
Quick Start
npx -y @stranzwersweb2/skillsync-mcp
Add to ~/.claude/settings.json:
{
"mcpServers": {
"skillsmp": {
"command": "npx",
"args": ["-y", "@stranzwersweb2/skillsync-mcp"]
}
}
}
Supported Clients
- Claude Code -
~/.claude/settings.json - OpenClaw -
~/.openclaw/mcp.json - Cursor -
.cursor/mcp.json - Windsurf -
~/.windsurf/mcp.json - GitHub Copilot -
.github/copilot-mcp.json - Zed -
~/.config/zed/settings.json - nanobot - MCP config
Security Model
| Risk Level | Behavior |
|---|---|
| Safe / Low | Installs with warnings shown |
| Medium / High | Blocked - requires force=true to override |
| Critical | Permanently blocked - no override possible |
Additional Safety Guards
- Path traversal prevention on skill names and filenames
- SSRF prevention - only
github.comURLs accepted npm install --ignore-scriptsblocks postinstall attacks- Max 50 files, 512KB per file, 2MB total size limit
- Output sanitization strips zero-width Unicode and bidi overrides
- 15-minute scan cache with 100-entry LRU eviction
- fs.watch startup verification with debounced re-scanning
Links
Author: Aditya Sugandhi | License: MIT | Version: v1.3.0
Server Config
{
"mcpServers": {
"skillsmp": {
"command": "npx",
"args": [
"-y",
"@stranzwersweb2/skillsync-mcp"
]
}
}
}