Server exposes 5 stock/finance tools with reasonable naming (verb-first) and bilingual descriptions. Schemas are present and mostly complete with enums and constraints. However, descriptions are thin (50-150 chars), parameter descriptions are minimal, output schemas are not documented, and error handling is generic. The server lacks actionable recovery guidance and does not segment error types for the LLM. Token return in error responses is a minor concern. Overall: above the community median but with clear gaps in LLM-optimization and error design.

• Document output schemas for all 5 tools. For get_stock_quote, specify: {name, code, price, change_percent, volume, turnover_rate, pe_ratio, ...}. For search_stock, specify array structure: [{code, name, type, ...}]. Use JSON Schema or TypeScript interfaces.
• Replace generic error handling with typed error responses. Instead of 'Error calling get_stock_quote: ...', return: {error: 'INVALID_CODE', code: '999999', message: 'Stock code must be 6 digits', retryable: false, suggestion: 'Use search_stock(keyword) to find the code'}. Classify errors as retryable (network) vs user-fixable (invalid code) vs fatal (API deprecated).
• Expand tool descriptions to 100 - 200 chars and answer: what does it do, when to use it, what does it return, any required preconditions. Example: 'Get real-time A-share stock quote (price, change%, volume, PE). Call this after search_stock(keyword) to get live data for a specific code. Returns current market data with ~15min delay from Eastmoney.'
• Add validation feedback for invalid 'code' parameter. Instead of returning a generic error, detect if code is malformed and suggest: 'Invalid code "9999". Expected 6 digits (e.g., 600519 for 贵州茅台). Try search_stock(keyword) first.'
• Document the rate-limiting and data freshness characteristics in descriptions. E.g., 'Data from Eastmoney with ~15min delay. Free tier: ~100 req/min. Premium data sources available via official API.'
• Add a discovery tool: describe_available_markets() returning {sh: Shanghai, sz: Shenzhen, cyb: ChiNext, kcb: STAR}. This helps LLMs understand market enum values without guessing.