Disclosed weakness · no CVE recorded

Claude Code cloud sessions could omit named connector deny rules

Anthropic · Claude Code · cloud and self-hosted runner sessions

First documented disclosure · Review checkpoint

Summary

Connector deny entries using the claude.ai name form did not apply consistently in cloud and self-hosted runner sessions.

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 release expands deniedMcpServers matching for the claude.ai connector-name form to these session types.

Conditions required

An administrator relies on a deniedMcpServers entry using claude.ai followed by the connector name.

Potential impact

The named connector could remain available where its configured deny entry was expected to apply.

Affected versions and configurations

The reviewed sources do not establish a complete affected-release range.

Fix and mitigation status

Anthropic lists the scope correction in 2.1.328.

Practical next steps

Upgrade to 2.1.328 and verify that named denied connectors are unavailable in cloud and self-hosted sessions.

Operating lesson

Apply administrator deny rules consistently across local and hosted execution environments.

Evidence and limits

The issue-specific public evidence is Anthropic’s November 19 release note. The proprietary implementation and a complete affected-version range were not independently established. General new hook controls and convenience features from the same release are not treated as separate vulnerabilities.

Sources