Four tools with complete schemas and descriptions, but descriptions are generic and lack LLM-optimized guidance. All tools are read-only with no state-modifying operations. Parameter descriptions are present but minimal (10-30 chars). No error handling guidance, no recovery paths, and no examples of chaining IDs across tools. Naming is clear and verb-prefixed. Output schemas are implicit but not formally documented. Missing: actionable error messages, parameter constraints (enums for mailbox/sender), and composition guidance.

• Expand parameter descriptions to 50-100 chars. For 'mailbox', write: 'Exact mailbox name to filter results. Call list_mailboxes() first to see available options (e.g., "Inbox", "Sent", "Archive").'
• Document output schemas in tool descriptions. E.g., 'Returns {count: int, results: [{id, subject, sender, sent_at, mailbox, preview?}]}' so LLMs know what fields to extract.
• Add enum constraints or validation hints. For search_messages, note: 'Valid mailbox values: Inbox, Sent, Archive, Drafts. Invalid values return empty results.'
• Include recovery guidance in error paths. When get_message returns {found: False}, suggest: 'Message not found. Try search_messages(query=...) to locate it by subject or sender.'
• Enhance tool descriptions with use-case examples. E.g., 'Use search_messages to find emails by keyword, then call get_message(message_id) to read the full body.'