Server has 7 well-named tools with consistent verb_noun patterns (shopify_sales_*, shopify_orders, shopify_customer_metrics, shopify_traffic). All tools have descriptions (avg 95 chars, within 10-1024 range). Input schemas are present with Zod validation and parameter descriptions. However, output schemas are NOT documented, responses return generic text content without structured field definitions. Error handling is absent; no recovery guidance or categorization. No tool annotations (readOnlyHint, etc.). Parameters lack constraints like min/max for numeric fields and explicit format specs. Missing pagination support despite tools returning potentially large datasets.

• Document output schemas for each tool. Define the structure of the text response (e.g., 'Returns a table with columns: product_title, total_sales, net_sales, orders, quantity'). Use structured JSON responses instead of plain text when possible.
• Add input validation and error handling. Validate date formats, enforce limit bounds (1-50 for first, 1-100 for limit), and return actionable error messages like 'Invalid date format: expected YYYY-MM-DD or -30d, got "invalid"'.
• Add tool annotations. Mark all tools with readOnlyHint: true since they only query metrics. This signals to LLMs that retries are safe.
• Implement pagination. Add offset/limit or cursor-based pagination to tools returning lists. Return total_count and next_cursor in responses so agents can fetch additional pages.
• Add parameter constraints to Zod schemas. Use .min(1).max(50) for numeric bounds and .regex() for date format validation. Update descriptions to reflect constraints (e.g., 'Number of orders to fetch (1-50, default 10)').