Server defines 11 tools with explicit schemas and descriptions in Chinese. All tools follow verb_noun naming (get_*). However, descriptions are minimal (10-30 chars), lack context on WHEN to use each tool, and provide no guidance on tool relationships or dependencies. Parameters are typed but lack detailed constraints. No output schemas documented. Error handling is generic ('錯誤：{str(e)}') with no recovery guidance. Tools are read-only but lack explicit idempotentHint annotations. No pagination support despite tools returning potentially large datasets (e.g., financials, revenue trends). Descriptions are too brief to meet the 50-200 char LLM-optimized baseline.

• Expand each tool description to 80-150 characters. Include WHAT it does, WHEN to use it (vs similar tools), and what it returns. Example: 'get_stock_signal: Returns buy/sell signals based on technical indicators (RSI, MACD, KD). Use this after get_stock_technical to get a composite recommendation. Returns signal (buy|sell|hold) and confidence score.'
• Document output schemas for all tools. Specify which fields are returned, their types, and which are required. Example for get_stock_info: {current_price: number, change_percent: number, pe_ratio: number, dividend_yield: number, market_cap: number}.
• Add pagination parameters (limit, offset) to tools returning lists (get_financials, get_revenue, get_stock_technical). Cap results at 20-50 items by default. Document in description: 'Returns up to 20 records; use offset to paginate.'
• Improve error handling. Catch specific exceptions (requests.Timeout, requests.ConnectionError, ValueError) and return actionable messages. Example: 'Stock ID not found. Valid format: 4-digit number (e.g., 2330). Try search_stocks() to find the correct ID.' Use error classification (retryable vs user-fixable).
• Add tool annotations to inputSchema. Mark all tools with readOnlyHint=true and idempotentHint=true. Example: Tool(..., inputSchema={...}, readOnlyHint=true, idempotentHint=true).