OS Command Injection Vulnerability in Anthropic Claude Code CLI and Claude Agent SDK
CVE-2026-35021

8.4HIGH

Key Information:

Vendor

Anthropic

Status
Claude Code
Claude Agent Sdk For Python
Vendor
CVE Published:
6 April 2026

Badges

👾 Exploit Exists

What is CVE-2026-35021?

The Anthropic Claude Code CLI and Claude Agent SDK expose a significant vulnerability that enables OS command injection through the prompt editor utility. By manipulating file paths with malicious input, attackers can exploit shell metacharacters like $() or backticks. These crafted file paths, despite being quoted, can still lead to command substitution and arbitrary command execution with the permissions of the invoking user. This vulnerability highlights the importance of proper input validation and sanitization to prevent unauthorized command execution.

Affected Version(s)

Claude Agent SDK for Python 0 <= 0.1.55

Claude Code 0 <= 2.1.91

References

https://phoenix.security/critical-ci-cd-nightmare-3-comma...
technical-descriptionexploit
https://www.vulncheck.com/advisories/anthropic-claude-cod...
third-party-advisory

CVSS V4

Score:
8.4
Severity:
HIGH
Confidentiality:
High
Integrity:
High
Availability:
High
Attack Vector:
Local
Attack Complexity:
Low
Attack Required:
None
Privileges Required:
Undefined
User Interaction:
Unknown

Timeline

  • 🟡

    Public PoC available

  • 👾

    Exploit known to exist

  • Vulnerability published

  • Vulnerability Reserved

Credit

Francesco Cipollone
.