Fixes #1451

Ensure HarnessAgent tool execution timeout exceeds task_output blocking waits and include tool identifiers in timeout errors.

AgentScope-Java Version

1.1.0-SNAPSHOT

Description

Background

When HarnessAgent enables subagent orchestration, task_output supports blocking up to 10 minutes via its timeout parameter (max: 600000 ms). However, the default tool execution configuration (ExecutionConfig.TOOL_DEFAULTS) applies a 5-minute timeout at the framework level.

As a result, a long-running task_output(block=true, timeout=600000) call can be interrupted by the tool executor before the tool's own wait limit is reached.

Changes

1. TaskTool

   • Introduce shared constants MAX_BLOCK_WAIT_MILLIS / MAX_BLOCK_WAIT (10 minutes) and reuse them in task_output timeout clamping.

2. HarnessAgent

   • When subagents are enabled, raise the effective tool execution timeout to TaskTool.MAX_BLOCK_WAIT + 1 minute (11 minutes) if the user config is unset or not greater than 10 minutes.
   • Apply subagent hook registration and timeout adjustment in the same code path.

3. ToolExecutor

   • Improve timeout error messages to include tool name and tool call id, e.g.
     Tool 'task_output' (id=...) execution timeout after PT5M.

Checklist

Please check the following items before code is ready to be reviewed.

• Code has been formatted with mvn spotless:apply
• All tests are passing (mvn test)
• Javadoc comments are complete and follow project conventions
• Related documentation has been updated (e.g. links, examples, etc.)
• Code is ready for review