Improper Access Control

@anthropic-ai/claude-code is vulnerable to improper access control. The vulnerability is due to an error in sed command parsing that bypasses read-only validation, which allows an attacker to write to arbitrary files on the host system...