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 influence authentication settings can inject shell metacharacters through parameters like apiKeyHelper, awsAuthRefresh, awsCredentialExport, and gcpAuthRefresh, leading to arbitrary command execution with the privileges of the user or automation environment. This can result in credential theft and exfiltration of environment variables. The vulnerability has a critical CVSS 4.0 score of 9.3. The affected product is a cloud service, and a patch is available, though no specific vendor advisory details were provided.

Successful exploitation allows an attacker to execute arbitrary OS commands with the privileges of the user or automation environment running the vulnerable components. This can lead to credential theft and exfiltration of sensitive environment variables, potentially compromising system integrity and confidentiality. The vulnerability is critical given the ease of exploitation (no privileges or user interaction required) and the high impact on confidentiality and integrity.

Since the product is a cloud service, the vendor typically manages remediation server-side. A patch is available for this vulnerability. Users should apply the official patch or update provided by Anthropic as soon as possible. In the absence of detailed vendor advisory content, check Anthropic's official channels for the latest remediation guidance. No additional mitigation steps are specified or required beyond applying the patch.