We need to use and explore Agentic AI to learn all the potentials and risks

Terok addresses the gap that Agentic AI for coding is so much more powerful than the previous "chatbot" mode but we (as humanity) don't know how to properly handle it, yet. It is risky because if the agent can run anything on your laptop that you could run, then things that go wrong may have hard consequences -- either due to misfortune or stupidity by the agent or due to evil intentions by someone. The basic precaution of asking the user for permission for every command no good solution because it would essentially limit the agent to "chatbot" mode again. But we need to explore the potential and the risks and we should not leave this to the carefree and unconcerned.

AI Agents in a Sandbox

Terok runs each agent task inside a hardened, rootless container with default-deny outbound networking, a credential vault that keeps real keys on the host, a per-task git checkpoint, and a desktop notification path for live allow/deny decisions. It ships a CLI and a Textual TUI on top of a focused stack of independently-released Python packages.

This sandbox environment for the AI agents allows to use agentic AI for coding in a safer and more secure way. It is work in progress and certainly not the final answer to the multiple challenges this new AI tool brings us. It is a solid step forward to a responsible way compared to the naïve manner.