Portfolio Daily Tracker exposes 14 tools with mixed quality. Naming is generally clear (verb_noun pattern: get_*, research_*, propose_*), but descriptions vary widely. Research tools (1-7) have minimal descriptions (10-60 chars, below the 194-char baseline). Market/portfolio tools (8-14) have better descriptions but lack parameter details. All tools have proper JSON schemas with type definitions. Critical gap: no output schemas documented, no error handling guidance, no pagination for list tools. Tool annotations present but underutilized. Chinese descriptions (tools 8-12) are descriptive but inconsistent with English tools.

• Expand research_* tool descriptions to 100-200 chars. Example: 'research_capabilities' → 'Inspect research limits and available model integrations. Call this first to understand which models and datasets are available before submitting a research_run.'
• Document output schemas for all tools. Add a 'Returns' section to each tool docstring listing field names, types, and meanings. Example: get_stock_quote returns {symbol, price, change_pct, volume, timestamp}.
• Add pagination to list tools: research_runs(limit=20, offset=0), get_market_news(limit=6, offset=0). Return total_count and next_offset in responses.
• Define the event object schema for propose_ledger_events. Document required fields (date, account, currency, quantity, price) and valid event types (buy, sell, dividend, fee). Use JSON Schema $ref or inline object definition.
• Add error recovery hints to tool descriptions. Example: get_stock_quote → 'If symbol not found, try compare_stocks() with a partial symbol to discover valid codes.'