31 tools with schemas present but quality is inconsistent. Most tools have descriptions (good), but many are in Chinese and lack LLM-optimized structure. Parameter descriptions exist but are often minimal (10-30 chars). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Output schemas are not documented. Error handling is minimal, no recovery guidance or actionable error messages visible. Naming is generally good (verb_noun pattern), but some tools combine multiple concerns (e.g., optimize_strategy handles both grid and bayesian modes). No evidence of input validation or constraint documentation beyond enum hints in descriptions.

• Add tool annotations: mark all read-only tools with readOnlyHint=true, backtest/optimize tools with idempotentHint=true (same params = same result), and any state-modifying tools with destructiveHint=true.
• Document output schemas for all tools. For backtest_strategy, define: {type: object, properties: {return_rate: {type: number}, sharpe_ratio: {type: number}, max_drawdown: {type: number}, trades: {type: array}}}.
• Translate descriptions to English and restructure as: '[ACTION] [WHEN/USE CASE] [RETURNS]. E.g., 'Get real-time stock quotes for A-shares, HK stocks, US stocks, or futures. Use this to fetch current price, volume, and change % before analysis. Returns: {code, price, change_pct, volume}.'
• Add error handling with recovery guidance. E.g., if market='invalid', return: 'Invalid market. Must be one of: a (A-shares), hk (HK stocks), us (US stocks), futures. Try again with a valid market.'
• Enforce JSON Schema constraints: add 'enum' for market/strategy/ktype params, 'pattern' for date strings (YYYY-MM-DD), 'minimum'/'maximum' for numeric ranges (e.g., limit: {type: integer, minimum: 1, maximum: 1000}).

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry