I know that you like to believe that AI agents have planning capabilities, but I would refer you to read the work of Prof. Subbarao Kambhapati. He has categorically shown that curretn AI systems can't plan.

Moreover, let me tell you what plan is, untill and unless a plan can't be formally verified, it is not a plan and that's not something AI agents can do.

I'm fully aware that they can be used with formal verifiers and in that sene they can plan, but not on their own. Verifying the correctness of a plan is a necessary condition to call it a plan.

Subbarao showed in so many of his research papers that, even the best of LLMs performs close to 0 in randmoized block world planning problem. You can read my blog on this as well; https://medium.com/aiguys/can-llms-really-reason-and-plan-50b0ac6addd8