Claude Code Tells Claude to Retry Narrower or Ask You After Safeguards Stop a Subagent
Claude CodeView original changelog
Claude CodeClaude Code 2.1.319 changes what Claude is told after safeguards stop a subagent. It is now told to retry with a narrower prompt or tell the user, instead of doing the work itself.
Key Takeaways
- Claude is now told to retry narrower or tell the user after a safeguard stops a subagent.
- It is no longer told to do the work itself.
- The change alters the main agent's next step.
- A narrower prompt is the suggested retry.
- The user stays in the loop under the new wording.
- The release notes describe the wording change and state no rationale.
A Changed Instruction
Claude Code 2.1.319 changes the message Claude receives when safeguards stop a subagent. Claude is now told to retry with a narrower prompt or tell the user. Before, the instruction let Claude do the work itself.
What Changes in Practice
When a safeguard stops a subagent, the main agent previously had a path to carry out the same work directly. The new instruction replaces that path with a narrower retry or a hand-back to the user. The release notes describe the wording change itself and do not state a rationale.