Single tool 'quote_request' has critical definition gaps. Input schema visible but severely underdeveloped: nested 'quote' object with no property definitions, no type constraints, no field descriptions. Description is 79 chars, acceptable length but lacks WHEN/WHY context and does not explain the WRITE risk or confirmation workflow. No output schema documented. Tool definition inferred from langsmith_evals.py (not primary server code), capping per-tool score at 50 max. STDIO transport further limits practical utility.

• Expand input schema: define 'quote' object properties (shipper, consignee, weight, dimensions, service_type, etc.) with types, descriptions, and constraints (enums for service_type, min/max for weight).
• Document output schema: specify response fields (confirmation_id, status, extracted_details, validation_errors) with types and descriptions.
• Enhance description: explain that this tool extracts and validates freight quote details from email, when to call it (after email ingestion), and what happens on success/failure.
• Add parameter descriptions: document what each 'quote' field represents, valid ranges, and format constraints (e.g. 'weight in kg, 0.1 - 50000').
• Implement error handling: return structured errors with actionable guidance (e.g. 'Missing shipper address. Provide: street, city, state, zip') and distinguish retryable vs fatal failures.