I see no point in this project. There ain’t any examples for the usage the author states his project is made for.

389 tokens saved? Ok. Since I pay per million tokens, what is the ratio here? Is there are any downside associated with output deletion?

Is Claude really using this behavior to make user bleed? I don’t think so.

PS: the author seems like a beginner. Agents feedback is always helpful so far and it also is part of inter agent communication. The author seems to lack experience.

As a lead I would not allow this to be included until proven otherwise: A/B testing.