What
Before Claude runs a shell command, Claude Code reads it to work out what it will do and whether it needs your permission. In a shell, a backslash at the end of a line joins it to the next line. That reader now follows such joins even when they fall in the middle of:
- operators such as
&& - redirections such as
<< - heredoc markers (a heredoc is a block of text passed to a command, ended by a chosen word)
- the openers
$(and${, which run or expand something inside a command
It also finds heredoc end words that are split across joined lines. The reader now marks a command as unreadable when it contains:
- a heredoc end word it cannot spell the way bash would
- a heredoc end word that reads as empty
- a
{that could either start a group of commands or be a plain argument - a
${...@P}expansion, which runs whatever the variable holds
Why
Before, a line break in the middle of &&, <<, $( or ${ could make the reader see a different command from the one the shell runs, which could hide a command inside it from permission checks. Now, when the reader cannot follow a command, it says so instead of trusting its own reading.
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubt
What happens next with a command marked unreadable is not settled, though most likely Claude Code asks you for permission.