AuthorProfileMCP has two tools with basic structure but significant quality gaps. Tool names follow the verb_noun convention (get_coauthors, get_author_keywords), which is good. However, parameter descriptions are minimal or missing, output schemas are not documented, and error handling is generic. The server catches exceptions but returns unstructured error responses that don't guide recovery. Parameter validation is absent, inputs accept any string with no format constraints, ranges, or enums. Descriptions are adequate (100+ chars each) but lack dependency hints or selection context. Tool responses return dictionaries with 'success' flags and raw error strings rather than actionable recovery guidance. The implementation shows awareness of async patterns and basic HTTP transport setup, but lacks production-grade tool quality standards.

• Document output schemas in tool definitions. Declare that get_coauthors returns {success: bool, author: str, institution: str, total_coauthors: int, coauthors: [{name: str, institution?: str, field?: str, ...}]}. Declare get_author_keywords returns {success: bool, author: str, source: str, total_keywords: int, keywords: [str]}. Schemas guide LLM reasoning and enable field extraction.
• Replace generic error handling with actionable recovery guidance. Instead of {'success': False, 'error': 'Connection timeout'}, return {'success': False, 'error_code': 'TIMEOUT', 'error_message': 'Google Scholar search timed out after 30s. Try again in 1 minute or use a more specific author name.', 'retry_after': 60}. Include the specific constraint violated and next steps.
• Add input constraints and validation. Define institution as a free-form string (50-200 chars, no special chars except hyphens/spaces). Define field as an enum or pattern (e.g., 'Computer Science', 'Biology', 'Physics'). Add min/max length for name/surname (2-50 chars each). Validate before querying APIs.
• Enhance parameter descriptions with format guidance. Update institution to: 'University or institution name (e.g., Stanford University, MIT). Used to disambiguate authors with common names.' Update field to: 'Research field or discipline (optional, examples: Computer Science, Machine Learning, Biology). Narrows search to relevant author profiles.'
• Implement result limits and pagination. Cap coauthors and keywords at 20-50 per response. Add optional limit and offset parameters to tool definitions. Document in descriptions: 'Returns up to 50 coauthors; use offset for pagination.' Return a next_cursor or has_more flag to guide the LLM.

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

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