MCP Management Skill

Discover and execute tools from MCP servers without polluting your context window.

When to Use

• Discovering capabilities: List available MCP tools/prompts/resources
• Task-based selection: Find relevant tools for specific operations
• Programmatic execution: Call MCP tools with proper parameter handling
• Context efficiency: Delegate MCP operations to subagents

Key Capabilities

Common Use Cases

Browse Available MCP Tools

Who: Developer exploring MCP ecosystem

"List all MCP tools configured in this project and save them for quick reference"

Execute Screenshot Tool

Who: QA engineer testing UI

"Take a screenshot of https://staging.example.com and save it to the screenshots folder"

Query Documentation

Who: Developer researching API

"Use context7 to find Next.js 15 app router documentation on data fetching patterns"

Store Entity Relationships

Who: Backend developer modeling data

"Create memory entities for User, Post, and Comment models with their relationships"

Coordinate Multiple Tools

Who: DevOps engineer automating tasks

"Take a screenshot of the admin panel, then use memory to store the current deployment state"

Pro Tips

Execution Priority:

1. Gemini CLI (primary) - Fast, automatic tool discovery
2. Direct scripts (secondary) - Manual tool specification
3. mcp-manager agent (fallback) - Context-efficient delegation

Critical: Use stdin piping for Gemini CLI, NOT -p flag (skips MCP initialization):

# ✅ Correct
echo "Take screenshot of example.com" | gemini -y -m gemini-2.5-flash

# ❌ Wrong - skips MCP init
gemini -p "Take screenshot" -y -m gemini-2.5-flash

GEMINI.md Format: Enforce structured JSON responses by adding:

"Return JSON only per GEMINI.md instructions"

Tool Catalog: Run list-tools to persist capabilities to JSON for fast offline reference.

Not activating? Say: “Use mcp-management skill to discover available MCP tools”

Configuration

MCP servers configured in .claude/.mcp.json:

{
  "mcpServers": {
    "context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp"]
    },
    "memory": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-memory"]
    }
  }
}

Gemini CLI Integration (recommended):

mkdir -p .gemini && ln -sf .claude/.mcp.json .gemini/settings.json

Quick Start

Method 1: Gemini CLI (Recommended)

# Install
npm install -g @google/gemini-cli

# Create symlink for shared config
mkdir -p .gemini && ln -sf .claude/.mcp.json .gemini/settings.json

# Execute with stdin piping
echo "Take a screenshot of https://example.com. Return JSON only per GEMINI.md instructions." | gemini -y -m gemini-2.5-flash

Expected Output:

{"server":"puppeteer","tool":"screenshot","success":true,"result":"screenshot.png","error":null}

Method 2: Direct Scripts

cd .claude/skills/mcp-management/scripts && npm install

# List tools (saves to assets/tools.json)
npx tsx cli.ts list-tools

# Execute tool
npx tsx cli.ts call-tool memory create_entities '{"entities":[{"name":"User","type":"model"}]}'

Method 3: mcp-manager Subagent

Use when Gemini CLI unavailable. Subagent handles discovery and execution, keeping main context clean.

Related Skills

• Chrome DevTools - Browser automation via MCP
• Problem Solving - Systematic debugging patterns
• Debugging - Error investigation strategies

Key Takeaway

MCP Management provides context-efficient access to external tools through Gemini CLI (primary), direct scripts (secondary), or subagent delegation (fallback). Persistent tool catalogs and structured JSON responses enable programmatic integration without context pollution.