This server has severe structural and quality issues. Tool definitions are scattered across multiple files with significant duplication (tools 1-11 include 4 exact duplicates: get_weather appears twice, execute_sql twice, process_refund twice, fetch_available_plans twice). Most tool descriptions are under 50 characters and lack actionable context. Parameter descriptions are minimal (e.g., 'name of the city' for city param). No output schemas are documented. Error handling is absent, tools return raw strings or mock responses with no guidance for LLM recovery. The codebase uses @openai/agents (a client library, not an MCP server) and does not implement the MCP protocol at all. This is a local agent orchestration script, not an MCP server.

• Migrate from @openai/agents to a proper MCP server implementation (e.g., mcp-server-node or mcp-server-python). This codebase is a local agent orchestration script, not a remote tool provider.
• Remove all duplicate tool definitions. Keep one canonical version of get_weather, execute_sql, process_refund, and fetch_available_plans.
• Expand tool descriptions to 50 - 200 characters. Include WHAT the tool does, WHEN to use it, and WHAT it returns. Example: 'Fetches current weather for a city. Returns temperature in Celsius and weather condition. Use when user asks about weather.'
• Document output schemas for all tools. Specify the structure of returned data (e.g., {city: string, degree_c: number, condition: string} for get_weather).
• Add error handling with recovery guidance. Instead of returning 'done', return {success: true, refundId: '...'} or {success: false, error: 'Customer not found. Try search_customers() first.'}.