mcp-tools

Provides operational guidance for configuring secure, scalable MCP-based AI systems.

Updated Mar 26, 2026
One-click install
npx skills add https://github.com/christophevg/c3 --skill mcp-tools-christophevg
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-tools
Source: https://github.com/christophevg/c3/tree/main/skills/mcp-tools
Command: npx skills add https://github.com/christophevg/c3 --skill mcp-tools-christophevg

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the confusion around how to discover, name, and grant Claude Code agents access to MCP (Model Context Protocol) tools—especially when using custom sub-agents that do not automatically inherit MCP tool access.

Core Features & Use Cases

  • MCP tool naming convention: Establishes the exact tool prefix format based on MCP server names so you can reference tools correctly.
  • Sub-agent access configuration: Provides the practical workaround for explicitly listing MCP tool names in a sub-agent’s allowed_tools definition.
  • Tool discovery workflow: Shows multiple ways to list available MCP servers and derive tool names for the specific environment you are running.
  • Security guardrails: Highlights the risk of destructive actions (e.g., email send/delete) and encourages granting only the tools a sub-agent needs.

Use Case: You have an MCP server configured for email access and want a custom sub-agent to search and fetch emails for a specific task, while ensuring it cannot use unrelated tools.

Quick Start

When your agent needs MCP access, explicitly list the MCP tool names (using the correct mcp__<server>__<tool> naming) under the sub-agent’s allowed_tools.

Frequently Asked Questions about mcp-tools

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I configure Claude Code sub-agents to access MCP server tools?

To configure Claude Code sub-agents for MCP server tools, explicitly list the desired tool names using the mcp__<server>__<tool> naming convention within the sub-agent's allowed_tools definition. Sub-agents do not automatically inherit MCP tool access.

What is the correct naming convention for MCP tools in Claude Code?

The correct naming convention for MCP tools in Claude Code is mcp__<server>__<tool>. This prefix format is derived directly from the MCP server name, allowing you to reference and configure tools accurately.

How do I discover available MCP servers and their tools for my agent workflow?

You can discover available MCP servers and derive their tool names by listing the servers configured in your specific runtime environment. This tool discovery workflow ensures you reference the exact tools available.

How can I prevent custom sub-agents from executing destructive MCP operations?

To prevent destructive MCP operations, apply security guardrails by granting only the specific tools a sub-agent needs in its allowed_tools list. Explicitly restricting access avoids unintended actions like email deletion.

Why does my custom Claude Code agent fail to use connected MCP tools?

Custom Claude Code agents fail to use connected MCP tools because they do not automatically inherit MCP tool access. You must explicitly list the required MCP tool names in the sub-agent's allowed_tools configuration.