I'm having trouble with a problem where values aren't constant and change midway through the process.
Specifically,
Agenticworkflow has AI agents A, B, C, and D, each with a different role.
When agent A produces a result of [Z: true], agent C then uses the result of Z to branch its instructions.
However, when actually running Agenticworkflow, while 80% of the time the correct values are exchanged,
about 20% of the time, a different value from A's is used by C.
I would appreciate it if you could tell me if there are any rules (proper practices) for exchanging values between agents.
I thought that normally, short-term memory would be used and the correct values would be exchanged between AI agents.
Regarding the data linkage we're considering:
Session retention using gs.getSession().getClientData
→ Is this possible?
Persistence via external files
→ Is this possible?
