mcp-builderlisted

# MCP Server Development Guide ## Overview Create MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. The quality of an MCP server is measured by how well it enables LLMs to accomplish real-world tasks. ## 1. Red Flags (Anti-Rationalization) **STOP and READ THIS if you are thinking:** - "I'll skip the evaluation phase because I tested it manually" -> **WRONG**. Manual testing is not reproducible. You **MUST** create evaluations. - "I'll just list all API endpoints as tools without designing workflows" -> **WRONG**. You **MUST** balance API coverage with usability. - "I'll use generic tool names like 'get_data'" -> **WRONG**. Tool names **MUST** be specific and descriptive (e.g., `github_list_repos`). - "I'll write the server in a language other than TypeScript or Python" -> **WRONG**. You **MUST** use the approved stacks for best compatibility. --- ## 2. Execution Policy ### 2.1 Execution Mode - **Mode**: `hybrid` - **Why this mode**: The skill requires both prompting for MCP design and logic generation, as well as iterative execution for testing the built servers. ### 2.2 Script Contract - **Command(s)**: - Depending on the target stack, testing relies on standard runtime commands: - Node: `npx @modelcontextprotocol/inspector` - Python: `python -m mcp run` - **Inputs**: Source code files for the MCP server. - **Outputs**: Running stdio process or compiled build artifacts. - **Failure semantics