Account memory is information Claude keeps about you across conversations. Claude Code has a layer that hides that content by replacing it with a placeholder. Before, every message went through that layer the same way.
It now first checks whether account memory is in play:
- When it is not, Claude Code takes a lighter path.
- When it is, messages marked as memory are replaced with a placeholder.
- When Claude Code cannot tell, or the check fails, it treats memory as in play.
Responses from hooks (commands you configure to run at set moments) are now handled according to which kind of event they came from.
This narrows when memory content is hidden. If the signal for whether memory is in play is wrong, content could be hidden when it should not be, or shown when it should be hidden. Falling back to treating memory as in play leans toward hiding.