Executive brief
Claude Code is an AI-powered code execution tool that requires user approval before running any command for security. Due to a parsing error in the echo command handler, an attacker could craft malicious code that bypasses this safety confirmation prompt and executes arbitrary commands without user consent. An attacker with the ability to inject code into a Claude Code conversation could execute untrusted system commands, potentially compromising the user's system and data.
Technical details
A command parsing error in Claude Code's echo command implementation allowed attackers to bypass the mandatory user confirmation prompt for command execution. The vulnerability exists in versions prior to 1.0.20 and is classified as CWE-78 (Improper Neutralization of Special Elements used in an OS Command). Attack requires network vector with low attack complexity and passive user interaction (e.g., opening a malicious code snippet in Claude Code). Successful exploitation grants high impact to confidentiality, integrity, and availability. The patch was released in version 1.0.20; users have been automatically updated or forced to the latest version (1.0.24+) which is no longer vulnerable.
Affected products
- Anthropic Claude Code before 1.0.20
Timeline
- 2025-08-04: disclosed: GHSA-x56v-x2h6-7j34 published
- 2025-08-05: advisory: CVE-2025-54795 assigned
- 2025-08-04: patched: Fix released in version 1.0.20; auto-updates and forced upgrades deployed