CVE-2026-35022 is an OS command injection vulnerability (CWE-78) in Anthropic Claude Code CLI and Claude Agent SDK. The issue arises because authentication helper configuration values are executed with shell=true without proper input validation. Attackers able to manipulate authentication parameters such as apiKeyHelper, awsAuthRefresh, awsCredentialExport, and gcpAuthRefresh can inject shell metacharacters to execute arbitrary OS commands with user or automation environment privileges. This can lead to credential theft and environment variable exfiltration. The vulnerability affects version 0 of the product and has a CVSS 4.0 score of 9.3 (critical). The product is a cloud service, and remediation is managed by the vendor.

Successful exploitation allows attackers to execute arbitrary OS commands with the privileges of the user or automation environment running the vulnerable components. This can result in theft of credentials and exfiltration of environment variables, potentially compromising system security and confidentiality. The vulnerability is rated critical with a high CVSS score of 9.3. No known exploits in the wild have been reported so far.

Since the affected product is a cloud service, the vendor Anthropic manages remediation server-side. A patch is available, and users should verify with the vendor advisory for confirmation of applied fixes. There is no indication that users need to take additional action if the service is updated by the vendor. Patch status should be confirmed by consulting the official vendor advisory.