ThemeParks.wiki MCP has clear, action-verb naming (list_, get_) and well-structured schemas with proper types and required fields. Descriptions are present and contextual (avg ~180 chars), guiding LLMs on discovery flow (list_destinations first, then feed IDs to other tools). However, parameter descriptions are minimal (single short phrases), output schemas are undocumented, and error handling lacks recovery guidance. No pagination support despite potential for large result sets. Tool composition is sound but lacks idempotency guarantees and per-item error reporting.

• Document output schemas for all tools. Add a 'returns' field to each tool definition specifying field names, types, and structure. Example: get_wait_times returns {park: string, attractions: [{id, name, status, wait_minutes, single_rider, last_updated}]}.
• Expand parameter descriptions to include format constraints. Change 'A PARK entity id from list_destinations' to 'A PARK entity id from list_destinations (alphanumeric string, e.g. "magic-kingdom")' to guide LLM input.
• Add pagination to get_wait_times and get_schedule. Introduce optional limit (default 20, max 100) and offset parameters. Return total_count and next_offset in response to enable multi-page iteration.
• Enhance error responses with recovery guidance. Instead of {error: 404, message: 'Not found'}, return {error: 'entity_not_found', message: 'Park ID not found', suggestion: 'Call list_destinations() to discover valid park IDs'}.
• Document idempotency and timeout behavior. Add a note: 'All tools are read-only and idempotent. Timeouts default to 10s; retries are safe.' This clarifies retry semantics for agents.