TaskOutputTool: Reads task output

TaskOutputTool is used to read background task output.
Its most important value is unification:

• Shell task output
• Local agent output
• Remote agent output

So the main thread doesn't need to worry about underlying task type differences, only about the unified return structure.

Input definition:

const inputSchema = z.strictObject({
  task_id: z.string(),
  block: z.boolean().default(true),
  timeout: z.number().min(0).max(600000).default(30000),
})

Unified output object:

type TaskOutput = {
  task_id: string;
  task_type: TaskType;
  status: string;
  description: string;
  output: string;
}

The source code prompt explicitly states:

DEPRECATED: Prefer using the Read tool on the task's output file path instead.

In other words, it still exists, but Anthropic is already steering "reading task output" toward the more generic Read tool.

This is valuable for studying the source code, as you can see how Claude Code's runtime capabilities are gradually converging.

Main thread wants to check background result > TaskOutputTool > read task status > wait to complete if necessary > uniformly extract output > return to main thread

TaskOutputTool acts as the "unified read interface" in the background task chain, even if it may eventually be replaced by the more generic Read tool.