RCE vulnerability found in Claude Code

Researchers discovered a remote code execution vulnerability in Claude Code version 2.1.118. The issue lies in the claude-cli deeplink handler and improper command-line argument parsing.

An attacker could craft a malicious link that, when clicked, launched Claude Code and silently loaded a malicious settings file. Due to a parsing error, config parameters were read directly from the request body. This led to arbitrary code execution on the victim’s machine.

The threat was heightened by bypassing repository trust dialogs. If the link pointed to a trusted project, the utility ran stealthily. Developers patched the vulnerability in update 2.1.119. Users are advised to update the tool and inspect local config files for suspicious hooks.