finz provides 8 financial data tools with consistent naming (all get_* verbs) and reasonable descriptions (100-150 chars each). All tools have single-parameter input schemas with type and description. However, output schemas are entirely undocumented, the code returns complex nested dicts (e.g., get_ticker returns 8+ nested objects) with no schema specification visible to the LLM. Parameter descriptions are minimal (e.g., 'Stock ticker symbol (e.g., AAPL)', under 50 chars). No error handling guidance, no pagination support despite tools returning variable-length results (news, earnings), and no input validation beyond type hints. Tools are well-composed (each does one thing) but lack the depth of documentation expected for production use.

• Document output schemas for all tools. For get_ticker, specify: {ticker: string, shortName: string, main_info: {sector: string, currentPrice: number, ...}, valuation: {...}, ...}. Use JSON Schema format so LLM can parse and plan.
• Add pagination to get_news, get_earnings, get_growth_estimate. Accept optional limit (default 20, max 100) and offset/cursor parameters. Return total_count and next_cursor in response.
• Expand parameter descriptions to 50-100 chars. Example: 'Stock ticker symbol (1-5 uppercase letters, e.g., AAPL). Must be a valid US equity symbol. Invalid symbols will return an error.'
• Wrap yfinance calls in try-except blocks. Return structured error responses: {error: 'Ticker not found', ticker: 'INVALID', suggestion: 'Did you mean AAPL?'} instead of raw exceptions.
• Add input validation: reject empty symbols, symbols longer than 5 chars, or symbols with non-alphanumeric characters. Return 'Invalid ticker format: must be 1-5 uppercase letters' before calling yfinance.