CVE-2025-64755: high-severity vulnerability in anthropics claude-code
@anthropic-ai/claude-code has Sed Command Validation Bypass that Allows Arbitrary File Writes
Published · Updated
E-mail alert when it lands in KEV, gets a public PoC or its score goes up.
No sign of exploitation. No public exploitation artifact known so far.
Claude Code had a flaw in how it validated sed commands, allowing attackers to bypass safety protections and write files anywhere on the system. This could let malicious code modify or create files, potentially compromising the entire system.
A sed command parsing error in Claude Code versions prior to 2.0.31 allowed bypassing read-only file validation, enabling arbitrary file writes via crafted sed commands. An attacker with control over Claude Code input could exploit this OS command injection vector to modify system files or inject malicious code, with no authentication required beyond tool access.
In the same product, most dangerous first.