Claude Code 2.1.209

現在你在跑代理的背景任務時，又可以順暢使用切換模型等互動選單，不會再被莫名擋下。以前，一個設定得太嚴格的防護機制（guard，用來阻擋非預期操作的保安設定）不小心封殺了 /model 指令與其他對話框，導致你在跑 claude agents（Claude 同時開多個分身代理幫你做事的功能）背景任務（background sessions，在主畫面背後默默執行的工作）時無法叫出這些互動介面。現在這個管太寬的限制已經被撤銷，讓你恢復應有的操作權限。就像辦公室換了新的電子鎖，結果設定太嚴格，連正常上班的員工都被鎖在自己的會議室外面；現在門禁規定撤回原本的標準，你總算能正常推開門辦公了。

You can once again use interactive menus, like changing AI models, without them getting stuck while running agent tasks behind the scenes. Previously, an overly strict guard (a protective mechanism meant to prevent unwanted actions) accidentally blocked the /model command and other dialogs from appearing in claude agents (a feature where Claude opens multiple agent instances to work for you) background sessions (tasks running silently in the background). This meant you couldn't interact with these interfaces when you needed to. We've rolled back this overly broad restriction so everything works normally again. It's like an office security door that was set so strictly it locked regular employees out of their own meeting rooms; the rule has been reverted to its standard setting so you can actually push the door open and get to work.