Agent Runtime: 'exec' tool needs testing for long-running continuous output

[wangtsiao]

Task

The exec tool (unified exec command handler at crates/core/src/tools/unified_exec/) needs testing, specifically for:

1. Long-running commands — Commands that run continuously and produce ongoing output (e.g., tail -f, ping, dev servers)
2. Streaming output updates — Output should be continuously streamed to the UI as it arrives
3. Interrupt/cancel — Ability to stop a running exec command

What to test

• Long-running processes don't hang or block the agent
• Output is progressively delivered to the client
• Cancellation/interruption works properly
• Resource cleanup after command completes or is cancelled

Related components

• crates/core/src/tools/unified_exec/ — exec tool implementation
• crates/core/src/tools/handlers/exec_command.rs — command handler