Critical gaps across all dimensions. Tool definitions are inferred from benchmark scripts rather than explicit schema registration in source code. No visible input schemas, parameter descriptions, or output documentation. Descriptions are minimal (10-50 chars) and lack LLM-optimized guidance on WHEN to use each tool or WHAT it returns. No error handling patterns, no pagination support visible, no parameter validation rules. The server exposes 12 tools but provides almost no structured metadata to guide agent selection or composition.

• Add explicit JSON Schema input definitions for all 12 tools in src/tools/*.ts. Each schema must include type, description, and constraints (enum, minLength, pattern, minimum, maximum) for every parameter. Reference: pattern:tool.
• Expand tool descriptions to 50 - 200 characters. Include WHEN to use (e.g., 'Use this to find messages matching keywords'), WHAT it returns (e.g., 'Returns up to 20 messages with sender, subject, date'), and any prerequisites (e.g., 'Requires Mail.Read scope'). Reference: pattern:tool-description.
• Add pagination to list tools: mail_list_messages, calendar_list_events, drive_search_items, gateway_search_tools. Include limit (default 20, max 100), offset/cursor, and total_count in responses. Reference: pattern:paginated-result.
• Implement error handling with recovery guidance. Return structured errors: {code, message, retryable, suggestion}. E.g., 'User not found. Try search_users() with a partial name.' Reference: pattern:recovery-guide.
• Document output schemas for all tools. Specify which fields are returned, their types, and which are safe to chain to downstream tools (e.g., user_id from graph_get_me must match parameter name in other tools). Reference: pattern:tool.

Inferred effective spec: 2026-07-28+.

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained