CVE-2026-39861: Claude Code: Sandbox Escape via Symlink Following Allows Arbitrary File Write Outside Workspace

CVE-2026-39861 · Severity: critical · CVSS 10 · Published 2026-04-21

Claude Code is a code execution environment that sandboxes untrusted code to prevent it from accessing files outside a designated workspace. A vulnerability in symlink handling allows an attacker to bypass this sandbox by creating symlinks that point outside the workspace, then writing through those symlinks to arbitrary locations on the system. Combined with prompt injection attacks, this could lead to arbitrary code execution outside the sandbox.

Technologies: Anthropic Claude Code, @anthropic-ai/claude-code (npm). Vendors: Anthropic, npm.

Quick answers

Is CVE-2026-39861 exploited in the wild?
No. Junglewise Threat Intelligence has no report of CVE-2026-39861 being exploited in the wild.
How severe is CVE-2026-39861?
Critical severity, with a CVSS score of 10, published 21 April 2026.
Which products and versions are affected?
Anthropic Claude Code < 2.1.64; @anthropic-ai/claude-code (npm) < 2.1.64, SEMVER:introduced=0,fixed=2.1.64.
Is there a fix for CVE-2026-39861?
Yes. @anthropic-ai/claude-code (npm): upgrade to 2.1.64 or later. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.
How is CVE-2026-39861 attacked?
It is attacked over the network, in a software library.
Am I affected by CVE-2026-39861?
Check whether you run Anthropic Claude Code or @anthropic-ai/claude-code (npm) at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

How to fix

  • @anthropic-ai/claude-code (npm): upgrade to 2.1.64 or later
    npm install @anthropic-ai/claude-code@2.1.64

Technical details

The vulnerability is a path traversal/sandbox escape in Claude Code's file handling logic (CWE-22, CWE-61). The sandboxed process can create symlinks pointing to locations outside the workspace, which the unsandboxed Claude Code process subsequently follows without validation. When Claude Code writes to a path that passes through such a symlink, it bypasses workspace boundary checks and writes to the target location without user confirmation. Attack requires the ability to inject untrusted code into Claude Code's context (via prompt injection) to trigger sandbox execution. The vulnerability allows arbitrary file writes outside the sandbox, potentially enabling code execution. A fix was released in version 2.1.64 and auto-deployed to standard users.

Affected products

  • Anthropic Claude Code < 2.1.64

Identifiers

  • CVE-2026-39861
  • GHSA-vp62-r36r-9xqp

Timeline

  • 2026-04-21: disclosed
  • 2026-04-21: patched: Fixed in version 2.1.64

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.