agent mode command

[bernoussama]

Is your feature request related to a problem? Please describe.
it cant execute commands on its own

Describe the solution you'd like
agent command:

• when active:
  • agent run command/script get its output
    • if successful exit
    • if not try to fix command/script or ask for more context if it needs it and repeat until successful
      and it's better to add safe field in model output schema to only run command if safe.