10 tools with basic schemas and descriptions, but significant gaps in parameter documentation and error handling. Most tools have verb-noun naming (get_, solve_, search_) which is good. Descriptions range 30-200 chars, meeting baseline minimums. However, 7 of 10 tools lack parameter descriptions entirely (e.g., solve_hanoi's 'n' parameter has no description of valid range or constraints). Error handling is minimal, most tools return generic error dicts without recovery guidance. No tool annotations (readOnlyHint, destructiveHint) despite all being READ_ONLY. Output schemas are documented in docstrings but not formally in JSON Schema. Matrix tools show better parameter documentation than time/weather/search tools.

• Add parameter descriptions to all 10 tools. For solve_hanoi, document 'n: integer, number of disks (must be >= 1, typically 1-15 for practical use)'. For search_internet, add 'query: string, search terms (max 500 chars, e.g. "python machine learning")'. For get_weather, add 'city: string, city name (e.g. "New York", "London").'
• Implement actionable error messages. Instead of {'error': 'N must be >= 1'}, return {'error': 'Invalid n: got 0. Must be an integer >= 1. Example: n=3 solves Hanoi for 3 disks.'}. For matrix tools, catch ValueError separately and return {'error': 'Invalid matrix: ' + specific reason}.
• Add readOnlyHint=True to all @mcp.tool decorators in FastMCP. This signals to agents these tools are safe to call without confirmation, improving agent efficiency.
• Document output schemas formally. For eigen_decomposition, add a return type hint or docstring: 'Returns: {"eigenvalues": [float | {"real": float, "imag": float}], "eigenvectors": [[float | {"real": float, "imag": float}]]}'. For search_internet, document: 'Returns: {"results": [{"title": str, "url": str, "snippet": str}], "count": int}'.
• Add limit and offset parameters to search_internet. Default limit=10, max=50. Return {"results": [...], "total": int, "limit": 10, "offset": 0} to enable pagination and prevent context explosion.