Agent context should turn safer reuse into better questions.
The boundary is not only a warning label.
It is a guide.
Here is what is known.
Here is what is uncertain.
Here is what should be asked next.
That is safer than pretending the note is a conclusion.