What
Before running a shell command, Claude Code breaks it into its separate commands and can check each against your permission rules. Previously it did this part-by-part check only when the input split into more than one command. It now also does it when the input becomes a single command whose text differs from what was written.
Why
Commands that Claude Code rewrites while reading them are now checked piece by piece, which makes permission checks stricter.
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubt
It is not clear which commands end up as a single command with different text.