🎯 SkillSMP MCP Server

Intelligent Skill Management for AI Agents

Search • Discover • Vet • Tier • Auto-Load

✨ What is SkillSMP MCP Server?

An MCP (Model Context Protocol) server that connects AI agents to the SkillSMP marketplace. It provides intelligent skill discovery, security vetting, and automatic tiering to help agents find and use the right skills at the right time.

🔍 Smart Discovery

• AI-powered semantic search
• Context-aware recommendations
• Auto-load high-confidence skills

🛡️ Security First

• Pattern-based vulnerability scanning
• Risk level assessment
• Pre-install vetting

📊 Intelligent Tiering

• Multi-dimensional scoring
• Feedback-driven promotion
• Automatic demotion

🚀 Easy Integration

• 18 MCP tools
• Works with Claude, Cursor, Codex
• Zero-config setup

📦 Installation

# Install globally
npm install -g @luckybalabalaya/skillsmp-mcp-server

# Or use with npx
npx @luckybalabalaya/skillsmp-mcp-server

Quick Setup

Add to your MCP client configuration:

{
  "mcpServers": {
    "skillsmp": {
      "command": "npx",
      "args": ["@luckybalabalaya/skillsmp-mcp-server"],
      "env": {
        "SKILLSMP_API_KEY": "your-api-key"
      }
    }
  }
}

🛠️ MCP Tools

Core Tools

| Tool | Description | |:-----|:------------| | skillsmp_search | 🔍 Search skills by keywords, category, tags | | skillsmp_ai_search | 🤖 AI-powered semantic search with relevance scores | | skillsmp_preview | 👀 Preview skill metadata and content | | skillsmp_install | 📥 Install skill to specified namespace | | skillsmp_list | 📋 List installed skills with filters | | skillsmp_load | ⬆️ Load skill content into context | | skillsmp_unload | ⬇️ Unload skill from active context |

Security & Vetting

| Tool | Description | |:-----|:------------| | skillsmp_vet | 🔒 Security scan skill content for risks |

Tier Management

| Tool | Description | |:-----|:------------| | skillsmp_promote | ⬆️ Promote skill to higher tier | | skillsmp_demote | ⬇️ Demote skill to lower tier | | skillsmp_rate | ⭐ Rate a skill (helpful/not-helpful/harmful) | | skillsmp_stats | 📊 View usage statistics | | skillsmp_metrics | 📈 Get detailed metrics for a skill | | skillsmp_evaluate | 🎯 Get tier evaluation with scoring breakdown |

Discovery

| Tool | Description | |:-----|:------------| | skillsmp_discover | 🔎 AI-powered skill discovery based on context | | skillsmp_auto_discover | 🚀 Automatically discover and recommend skills | | skillsmp_add_context | 📝 Add message to discovery context |

🏆 Intelligent Tiering System

Skills progress through tiers based on multi-dimensional scoring, not just usage frequency:

┌─────────────────────────────────────────────────────────────┐
│                      TIER PROGRESSION                        │
├─────────────────────────────────────────────────────────────┤
│                                                              │
│   ┌──────────┐    ┌──────────┐    ┌──────────┐             │
│   │  SANDBOX │ ──▶│   WARM   │ ──▶│    HOT   │             │
│   │   🔸      │    │   🔶     │    │   🔥     │             │
│   └──────────┘    └──────────┘    └──────────┘             │
│       New           Regular          Frequent               │
│       Skills         Usage            Success               │
│                                                              │
└─────────────────────────────────────────────────────────────┘

Scoring Dimensions

| Metric | Weight | Description | |:-------|:------:|:------------| | 🎯 Relevance | 35% | How well skill matches current context | | ✅ Success Rate | 25% | Ratio of helpful vs not-helpful feedback | | ⏰ Recency | 20% | Recent usage weighted higher | | 🗂️ Project Match | 15% | Match with project languages/frameworks | | 📝 Content Quality | 5% | Skill metadata completeness |

🔐 Security Vetting

All skills are scanned before installation:

| Risk Level | Description | |:-----------|:------------| | 🟢 LOW | Safe to install | | 🟡 MEDIUM | Review recommended | | 🟠 HIGH | Manual review required | | 🔴 CRITICAL | Installation blocked |

Scanned for:

• 🔑 Secret/API key patterns
• ⚠️ Suspicious code patterns (eval, exec, etc.)
• 💀 Dangerous system commands
• 📦 Insecure dependencies

🎯 Skill Orchestrator (v1.4.0)

The new Skill Orchestrator provides intelligent skill automation:

User Message → Intent Detection → Skill Matching → Auto Load → Feedback Loop

Components

| Component | Description | |:----------|:------------| | IntentEngine | Analyzes user messages to detect intent (code generation, debugging, testing, etc.) | | TriggerConfigManager | Manages skill trigger rules (builtin, learned, user-defined) | | SkillMatcher | Matches intents to relevant skills using rules and AI search | | SkillRunner | Executes skill installation, loading, and unloading | | FeedbackLoop | Collects feedback, evaluates effectiveness, auto-adjusts rules |

Usage

import { SkillOrchestrator } from '@luckybalabalaya/skillsmp-mcp-server';

const orchestrator = new SkillOrchestrator(client, storage, metrics);

// Analyze user message and get skill recommendations
const result = await orchestrator.processMessage("Help me debug this TypeScript error");

// result.intent.type === 'debugging'
// result.matches === [{ skillId: 'debugging', confidence: 'high', ... }]
// result.loaded === [successful loads]

Learning from Feedback

The orchestrator learns from user feedback:

// Positive feedback - promotes skill and learns rule
await orchestrator.submitPositiveFeedback('debugging', 'Fixed the issue quickly');

// Negative feedback - demotes skill and adjusts rules
await orchestrator.submitNegativeFeedback('debugging', 'Not relevant for this context');

// Switch feedback - learns alternative skill
await orchestrator.submitSwitchFeedback('debugging', 'tdd-workflow', 'Actually need to write tests');

🤖 Auto-Discovery

The auto-discovery system analyzes conversation context and proactively recommends relevant skills:

// Add conversation context
skillsmp_add_context({
  role: "user",
  content: "I need to implement react hooks for state management"
})

// Trigger discovery
skillsmp_auto_discover({ force: true })

// Result:
// ✅ Detected: TypeScript, React
// 📚 Recommended: react-hooks-patterns (score: 0.92)
// 🚀 Auto-loaded: react-hooks-patterns

Auto-Load Thresholds

| Confidence | Threshold | Behavior | |:-----------|:----------|:---------| | high | 0.4 | Aggressive auto-loading | | medium | 0.6 | Balanced (default) | | low | 0.8 | Conservative |

⚙️ Configuration

# Required
SKILLSMP_API_KEY=sk_live_xxx           # Your SkillSMP API key
SKILLSMP_BASE_URL=https://skillsmp.com # API base URL

# Optional
SKILLSMP_CACHE_TTL=24                  # Cache TTL in hours
SKILLSMP_AUTO_VET=true                 # Auto-scan on install
SKILLSMP_DEFAULT_NAMESPACE=sandbox     # Default namespace
SKILLSMP_DISCOVERY_MODE=passive        # passive | active
SKILLSMP_DATA_DIR=~/skillsmp           # Data directory
SKILLSMP_AUTO_LOAD_THRESHOLD=medium    # low | medium | high

📁 Project Structure

skillsmp-mcp-server/
├── 📂 src/
│   ├── 📄 index.ts              # Entry point
│   ├── 📄 server.ts             # MCP server (18 tools)
│   ├── 📄 config.ts             # Configuration
│   ├── 📂 client/
│   │   └── 📄 skillsmp.ts       # SkillSMP API client
│   ├── 📂 storage/
│   │   └── 📄 manager.ts        # Storage & tier management
│   ├── 📂 metrics/
│   │   ├── 📄 tracker.ts        # Metrics tracking
│   │   └── 📄 types.ts          # Metrics types
│   ├── 📂 discovery/
│   │   ├── 📄 orchestrator.ts   # Discovery orchestration
│   │   ├── 📄 context-analyzer.ts
│   │   └── 📄 skill-recommender.ts
│   ├── 📂 vetting/
│   │   ├── 📄 scanner.ts        # Security scanner
│   │   └── 📄 risk-calculator.ts
│   └── 📂 types/
│       └── 📄 *.ts              # Type definitions
├── 📂 __tests__/
│   ├── 📄 storage.test.ts       # 12 tests
│   └── 📄 metrics.test.ts       # 12 tests
└── 📄 package.json

🧪 Development

# Install dependencies
npm install

# Development mode
npm run dev

# Build
npm run build

# Test
npm test

# Lint
npm run lint

📄 License

MIT © 2024

Made with ❤️ for AI Agents

🌐 SkillSMP • 📦 npm • 🐛 Issues