IntelHub exposes 4 tools with severe definition gaps. Tool names lack action verbs (initDrawTool, mountDrawTool, createApplicationTools, resolveGoogleServerKey are imperative but not verb_noun patterns). Descriptions exist but are verbose (100-200+ chars) and lack clarity on WHEN to use each tool or what state changes occur. Input schemas are present but parameters lack type definitions beyond 'object', no nested field documentation, no enums, no constraints. Output schemas are completely undocumented. No error handling guidance. Tools appear to be UI/rendering utilities rather than agent-callable operations, raising fundamental questions about whether this is a valid MCP server for agentic use.

• Rename tools to verb_noun pattern: init_draw_tool → enable_globe_drawing, mount_draw_tool → attach_draw_handler, create_application_tools → initialize_scene_tools, resolve_google_server_key → get_google_api_key. This makes intent immediately clear to LLMs.
• Document complete input schemas with nested field types. For 'viewer' param, specify: {type: 'object', properties: {scene: {type: 'object'}, pickPosition: {type: 'function'}}, required: ['scene', 'pickPosition']}. Every parameter must have a type, not just 'object'.
• Add output schemas to every tool. E.g., initDrawTool returns {type: 'object', properties: {handle: {type: 'object', properties: {destroy: {type: 'function'}}}, status: {type: 'string', enum: ['ready', 'error']}}}.
• Add error handling guidance: 'If viewer is null, return error: "Viewer not initialized. Call initialize_scene_tools first." If drawing fails, return {error: 'Drawing failed', retryable: true, reason: 'Scene not ready'}'.
• Reconsider whether these are agent tools or library functions. If they are library functions, expose them as resources or prompts instead. If they are agent tools, redesign them as discrete, stateless operations (e.g., 'get_drawing_state', 'submit_drawing_coordinates', 'finalize_drawing') rather than stateful UI handlers.