Claude Code Narrows Its Account Email Check to Commands That Set a User-Agent
Claude CodeClaude Code 2.1.321 changes the account email check so that it refuses only a shell command that contains the user's email and names a User-Agent, and it does so without asking. Other commands and scheduled runs no longer stop at its prompt.
Key Takeaways
- The email check now targets one pattern: a command containing the account email and a User-Agent.
- Matching commands are refused without asking, instead of raising a prompt.
- Other commands no longer stop at the prompt.
- Scheduled and unattended runs stop stalling on a question nobody can answer.
- The privacy goal is unchanged: the account email should not leave in a User-Agent.
- It refines an earlier change, in which Claude Code began asking before sending the account email to the network.
What the check does
Earlier, Claude Code added a check that stops it from sending a user's account email to the network without permission. Claude Code 2.1.321 narrows that check. It now refuses only a shell command that both holds the user's email and names a User-Agent, which is the pattern for identifying a client to an outside service. It refuses such a command outright instead of asking.
What improves
Before this change, the check could stop a command at a prompt even when the command was harmless. That was a problem for scheduled and unattended runs, which have no one present to answer. After the change, other commands run without that prompt, and scheduled runs no longer stall on it. The protection against leaking the account email into an outbound User-Agent remains.