Disclosed weakness · no CVE recorded

Claude Code subagents resumed by sibling messages could use mismatched permissions

Anthropic · Claude Code · resumed subagents

First documented disclosure · Review checkpoint

Summary

A subagent resumed by another subagent’s message did not consistently follow the session’s current permission settings.

Severity and exploitation evidence

No CVSS score recorded in this review.

Exploitation not established: The reviewed sources describe a correction or controlled reproduction; malicious exploitation against users is not established.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The resume path did not consistently apply the session permission state to the resumed agent.

Conditions required

A sibling message resumes an existing subagent after permission state has changed or requires a user decision.

Potential impact

The resumed work could proceed without the expected current-session permission decision.

Affected versions and configurations

The reviewed upstream report does not establish a complete affected-release range.

Fix and mitigation status

Anthropic lists the permission-state correction in 2.1.327.

Practical next steps

Upgrade to 2.1.327 and review resumed-agent behavior under the intended session permission policy.

Operating lesson

Reapply current authority when work resumes through a different trigger.

Evidence and limits

Anthropic’s release notes are the issue-specific public evidence; the proprietary implementation was not independently inspected. No CVE, score or full affected-version range is provided. The release does not enumerate the exact earlier permission combinations; impact is limited to the mismatch described by the maintainer.

Sources