Provide agent with internal /stop tool, the ability to `kill` spawned processes

What variant of Codex are you using?

CLI

What feature would you like to see?

The agent will often boot a process, then realise it has hung. The user can type /stop to kill the subprocesses easily. The agent will attempt to use ps -axo pid,ppid,command | rg "(...)", kill. If you're using the automatic sandbox checker, the supervisor LLM is needlessly asked if it's safe, (and doesn't seem to have the provenance information?). Since the process provenance is known by the hoist, the agent spawned the process in the first place, I don't see why the agent can't have access to a stop on its spawned processes?

Additional information

_No response_