Provide models with context about shell history

[ivy]

This would be a tough one to do without risk of leaking sensitive information but would go a long way towards providing context about what was done in between runs. I'm not sure yet if this would be feasible. 🤔