Server has 5 tools with basic schemas and descriptions in Japanese. Critical gaps: (1) get_customers has empty input schema ({}), scoring 0 for schema; (2) descriptions are non-English and very short (10-20 chars), below the 50-200 char LLM-optimized baseline; (3) no error handling guidance, errors return generic 'Failed to search customers' with no recovery hints; (4) no output schemas documented; (5) no parameter validation or constraints (e.g., per_page 1-50 stated in description but not enforced); (6) create_customer lacks confirmation/dry-run for destructive operation. Naming is acceptable (verb_noun pattern), but descriptions are too terse and non-English, limiting LLM comprehension.

• Rewrite all tool descriptions in English, 80-150 characters, following the pattern: 'Searches for customers by [criteria]. Returns customer ID, name, email, and phone. Use this to find existing contacts before creating or updating. Requires customer_group_id.'
• Add explicit output schema documentation for each tool. Example: 'Returns: {customers: [{id, name, email, phone, created_at}], total_count, page, per_page}'
• Implement input validation with actionable error messages. Example: 'Invalid per_page: got 100, must be 1-50. Invalid page: got 0, must be >= 1.'
• Add a confirm_create_customer tool or a dry_run parameter to create_customer to prevent accidental duplicates.
• Enhance error responses with recovery hints. Example: 'Customer not found. Try search_customers() with a partial name or email.' instead of 'Failed to search customers'.