Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. get_unread_emails has a simple, well-named interface but lacks output schema documentation. create_draft_reply has a clear name and docstring but the mutual exclusivity constraint (original_message_id OR thread_id) is documented in the description rather than enforced via schema or validation. Neither tool documents return types or field structures. Parameter descriptions are minimal (e.g., 'Plain-text reply content' for reply_body lacks format/length constraints). No error handling guidance, no pagination support for potentially large result sets, and no tool annotations (readOnlyHint/destructiveHint). Naming follows verb_noun convention well; descriptions exist but are brief (under 100 chars each). Schemas present but incomplete, no documented output structure, no validation rules, no recovery guidance.

• Document output schemas for both tools. Specify: get_unread_emails returns {emails: [{id, subject, from, snippet, timestamp}], total_count}, create_draft_reply returns {draft_id, thread_id, message_id}.
• Add parameter validation and error messages. If both original_message_id and thread_id are provided, return 'Provide either original_message_id or thread_id, not both.' If neither provided, return 'Must provide one of: original_message_id or thread_id.'
• Expand parameter descriptions with constraints. E.g., 'reply_body: Plain-text reply content (1-10000 characters, no HTML).' and 'limit: Maximum unread emails to return (1-100, default 10).'
• Add tool annotations via FastMCP. Mark get_unread_emails with readOnlyHint=true; mark create_draft_reply with destructiveHint=true (modifies Gmail state).
• Implement pagination for get_unread_emails. Add offset parameter, return {emails: [...], total_unread_count, has_more: bool} so agents can fetch additional batches.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry