POP: a tiny CrewAI-style trial for portable persona objects

Hi everyone,

I’m experimenting with a small prototype called POP (Persona Object Protocol).

The idea is to treat persona as a portable object layer instead of keeping persona scattered across prompts or config fragments.

I just added a small first-trial entry:

  • TRY_POP.md
  • a tiny dry-run team demo
  • 3 persona refs projected into 3 CrewAI-style runtime roles

Repo:

Demo:

The demo shows a simple team:

Designer
Engineer
Marketing Manager

All roles are loaded as persona objects.

I’m mainly looking for early feedback:

  • Is the entry path clear?
  • Does this approach make persona handling easier?
  • Would this be useful for CrewAI-style agent teams?

Any feedback is welcome.