The server defines 2 tools with HTTP transport via fastmcp. Both tools have descriptions and basic input schemas. Tool names follow verb_noun convention ('classify_occupation', 'get_income_statistics'). However, there are significant gaps: parameter descriptions are minimal (under 20 chars in most cases), output schemas are not documented in the code, error handling guidance is absent, and no tool annotations (readOnlyHint/destructiveHint) are present despite both tools being READ_ONLY. The SearchEngine.search() and SCBClient.get_income_statistics() implementations are not visible, making it impossible to verify actual return structure or error handling behavior. Input validation exists (title non-empty, description length capped at 4000 chars) but error messages are generic. Overall: solid naming and basic structure, but descriptions lack LLM-optimization detail and output contracts are undefined.

• Expand parameter descriptions to 50-100 characters with context. Example: 'The job title (e.g., "Software Engineer"). This is used to search the Swedish SSYK occupation taxonomy. If you only have partial information, the 'description' field can provide additional context like key responsibilities or industry.'
• Document output schemas explicitly in tool docstrings. Example for classify_occupation: 'Returns a list of matched SSYK records, each with: { ssyk_code (str): 4-digit Swedish occupation code, title (str): job title in Swedish, score (float): relevance score 0-1 }. Results are sorted by score descending.'
• Add error handling guidance to descriptions. Example: 'If the search returns no matches, try a shorter title or add a description. If the SSYK code is invalid, call classify_occupation to find the correct code first.'
• Add tool annotations to the fastmcp tool registration. Both tools should have @mcp.tool(readOnlyHint=True) or equivalent to signal they are safe for repeated calls.
• Document the search result limit and sorting order. Example: 'Returns up to 20 best-matching SSYK codes ranked by relevance score (highest first). If no matches found above threshold, returns empty list.'
• Specify the valid format and range for ssyk_code parameter. Example: 'The 4-digit SSYK code from the Swedish occupational classification system (format: digits only, e.g., "2512"). Invalid codes return an error with available matches.'
• Add type hints and descriptions to the response structure in the code. Use Pydantic models for classify_occupation (e.g., SSYKMatch with ssyk_code: str, title: str, score: float) so the schema is automatically documented.