I like this approach for Nix: https://dev.to/andersonjoseph/how-i-run-llm-agents-in-a-secure-nix-sandbox-1899 https://dev.to/andersonjoseph/how-i-run-llm-agents

I like this approach for Nix: https://dev.to/andersonjoseph/how-i-run-llm-agents-in-a-secure-nix-sandbox-1899 https://dev.to/andersonjoseph/how-i-run-llm-agents-in-a-secu... It makes it also easy to give the agent only access to the tools it actually needs.