feat: plan review/acceptance

[pkarw]

I can imagine that with the agents running some potentially risky operations (fs, shell, integrations) users might want to accept the execution plan before running it. I think it would be great to pass this desire as a teamwork exec option or something like this

[pkarw]

https://github.com/Significant-Gravitas/AutoGPT had sth like this in the early versions (and probably still have), same situation with https://github.com/OpenInterpreter/open-interpreter

[pkarw]

Having the execution plan pre-planned, we'd be able to add a second phase plan optimization (like multiple-phase compilers do) - merging some tasks, etc.