Single tool with clear naming (get_task_status) and good description (89 chars, action-oriented). Input schema is properly typed with Zod validation and includes minItems constraint. However, output schema is completely undocumented, the tool returns structured data but LLMs cannot see what fields to expect. No error handling guidance. Tool annotations present (readOnlyHint, openWorldHint) but incomplete. Missing parameter descriptions for task_ids array items and no guidance on what constitutes valid task IDs or expected response structure.

• Document output schema explicitly: 'Returns {success: boolean, tasks: [{id: integer, type: string, status: string, progress_percentage: number, actions_completed: integer, actions_required: integer}]}'. Include field descriptions and constraints.
• Enhance task_ids parameter description: 'Array of positive integers (1-999999) identifying tasks to check. Max 100 per call. Returns 404 if any ID not found; use list_tasks() to discover valid IDs.'
• Add error handling guidance to tool description: 'On failure, returns {success: false, http_status: number, error: string}. If 404, task does not exist, verify ID with list_tasks(). If 401, API key invalid.'
• Clarify use case in description: 'Fetches real-time progress for one or more tasks. Use this to poll task status during long-running operations. For bulk task discovery, use list_tasks() instead.'
• Add result limit constraint: 'Accepts 1 - 100 task IDs per call. Responses are paginated if needed. Returns at most 100 task records per request.'