Confirmation prompts should ideally come from gptscript and not LLM when executing commands.

[sangee2004]

gptscript version v0.0.0-dev-671f0f02-dirty

Steps to reproduce the problem:

1. Execute cli-demo script - gptscript --disable-cache github.com/gptscript-ai/cli-demo
2. Try to chat with k8s agent to get it to do execute any k8s command.
3. Notice that sometimes confirmation prompts comes from LLM when executing commands.

Expected Behavior
Confirmation prompts should ideally come from gptscript.

We see the LLM prompt for confirmation:

> scale up  nginx deployment to 5

  I will scale the  nginx  deployment to 5 replicas. Let's proceed.                                                                                                  
                                                                                                                                                                     
  Executing the command:                                                                                                                                             
                                                                                                                                                                     
    kubectl scale deployment nginx --replicas=5                                                                                                                      
                                                                                                                                                                     
  Shall I go ahead?                                                                                                                                                  

> no

  Alright, I won't proceed with scaling the deployment. If there's anything else you need, feel free to let me know!                                                 

It should ideally have come from gptscript

> scale up  nginx deployment to 5

    ┌────────────────────────────────────────────────────┐
    │ Call Arguments:                                    │
    │                                                    │
    │ k8s {"task":"scale deployment nginx --replicas=5"} │
    └────────────────────────────────────────────────────┘
                                                          
    ┌─────────────────┐
    │ Call Arguments: │
    │                 │
    │ exec {"         │
    └─────────────────┘
                       
Run "kubectl scale deployment nginx --replicas=5" (or allow all "kubectl scale ..." commands)
Confirm (y/n/a)>