Five tools with complete input schemas and detailed descriptions. Naming follows verb_noun pattern (ashares_*) and is domain-specific. Descriptions are comprehensive (150-250 chars), exceeding the 10-1024 baseline. However, output schemas are entirely undocumented, LLMs cannot predict response structure for downstream chaining. No error handling guidance, no parameter validation rules (e.g., date format flexibility, symbol format), and no indication of rate limits or data freshness. Tools are read-only (good for safety) but lack idempotency hints. Parameter descriptions are strong but lack explicit constraints (e.g., 'limit: 1 - 100' stated in description but not as JSON Schema minItems/maxItems).

• Document output schemas for all five tools. Example for ashares_quote: { type: 'object', properties: { symbol: { type: 'string' }, name: { type: 'string' }, price: { type: 'number' }, change_pct: { type: 'number' }, timestamp: { type: 'string', format: 'iso8601' } }, required: ['symbol', 'price'] }
• Add error recovery hints to parameter descriptions. E.g., ashares_quote: 'If symbol is unknown, call search_symbols() with a partial name or Chinese company name.'
• Formalize constraints in JSON Schema. Replace prose 'limit: 1 - 100' with { type: 'number', minimum: 1, maximum: 100, default: 30 }. Replace 'date: YYYYMMDD or YYYY-MM-DD' with { type: 'string', pattern: '^\d{4}-?\d{2}-?\d{2}$' }.
• Add readOnlyHint: true to all tool definitions to signal safety to agents. Example: { name: 'ashares_quote', description: '...', inputSchema: {...}, readOnlyHint: true }
• Document pagination and result limits. For ashares_limit_up, clarify: 'Returns up to `limit` stocks (default 30, max 100). Always includes total count of all limit-up stocks for the day. No pagination cursor needed, all results fit in one response.'