CVE-2026-33068: @anthropic ai/claude code Claude Code Has A Workspace Vulnerability

HIGH
Advisory Identifier: GHSA-mmgp-wc2j-qcv7
Share
CWE-807
Published
Advisory Summary

CVE-2026-33068 is a HIGH claude code has a workspace vulnerability (CWE-807: Reliance on Untrusted Inputs in a Security Decision) affecting @anthropic ai/claude code libraries in the npm ecosystem. The issue involves claude code has a workspace trust dialog bypass via repo-controlled settings file. Remediation is available by upgrading affected packages to version 2.1.53.

CVE-2026-33068 at a Glance

CVE Identifier CVE-2026-33068
GHSA Identifier GHSA-mmgp-wc2j-qcv7
Severity HIGH
CVSS v3.1 Base Score N/A
CWE Classification CWE-807: Reliance on Untrusted Inputs in a Security Decision
Affected Product @anthropic ai/claude code
Package Ecosystem npm
Published Date
Last Updated

How to Fix CVE-2026-33068

To remediate this vulnerability, upgrade all affected @anthropic ai/claude code dependencies to a patched release. Both direct and transitive dependencies should be checked, as frameworks or security libraries may pull in vulnerable versions transitively.

@anthropic-ai/claude-code fixed: 2.1.53 
npm install @anthropic-ai/claude-code@2.1.53
Verify Direct and Transitive Dependencies

Run the following command in your project directory to verify that all resolved dependencies use the patched release: 

npm ls @anthropic-ai/claude-code

How to Check if You Are Affected

To determine whether your Node.js application is affected by CVE-2026-33068, inspect your project dependencies using npm:

npm ls @anthropic-ai/claude-code

Analyze the terminal output to locate any version of @anthropic ai/claude code in your dependency tree. If any resolved version matches the vulnerable ranges listed in the table below (for example, versions earlier than 2.1.53), your application is affected.

Note: The npm ls @anthropic-ai/claude-code command inspects dependencies declared in your package.json and lockfile; it does not perform automated runtime dynamic scanning.

Technical Details

Vulnerability Mechanism & Technical Analysis

This vulnerability is classified under CWE-807: Reliance on Untrusted Inputs in a Security Decision, where the product uses a protection mechanism that relies on the existence or values of an input, but the input can be modified by an untrusted actor in a way that bypasses the protection mechanism. In @anthropic ai/claude code, software state validation or input boundaries are circumvented during processing. Review the official source advisory statement below for specific affected components, classes, and reproduction details.

Source Advisory Statement

Claude Code resolved the permission mode from settings files, including the repo-controlled `.claude/settings.json`, before determining whether to display the workspace trust confirmation dialog. A malicious repository could set `permissions.defaultMode` to `bypassPermissions` in its committed `.claude/settings.json`, causing the trust dialog to be silently skipped on first open. This allowed a user to be placed into a permissive mode without seeing the trust confirmation prompt, making it easier for an attacker-controlled repository to gain tool execution without explicit user consent. Users on standard Claude Code auto-update have received this fix already. Users performing manual updates are advised to update to the latest version. Thank you to hackerone.com/cantina_xyz for reporting this issue.

CVE-2026-33068 Affected @anthropic ai/claude code Versions

1 component(s)
Fixed Versions:
2.1.53
Product / Package Ecosystem Affected Versions Fixed Version
@anthropic-ai/claude-code
npm/@anthropic-ai/claude-code
npm
< 2.1.53
2.1.53

References

GitHub Advisory / Wiki ADVISORY
https://github.com/anthropics/claude-code/security/advisories/GHSA-mmgp-wc2j-qcv7
NVD — CVE-2026-33068 NVD
https://nvd.nist.gov/vuln/detail/CVE-2026-33068
@anthropic ai/claude code Source Repository REPOSITORY
https://github.com/anthropics/claude-code

Frequently Asked Questions

What is CVE-2026-33068?
CVE-2026-33068 is a security vulnerability in @anthropic ai/claude code involving Claude Code has a Workspace Trust Dialog Bypass via Repo-Controlled Settings File under CWE-807: Reliance on Untrusted Inputs in a Security Decision.
Which @anthropic ai/claude code versions are affected?
Affected versions include releases prior to 2.1.53 across affected packages: @anthropic-ai/claude-code.
What version fixes CVE-2026-33068?
Upgrading to 2.1.53 (or the latest patched release for your respective branch) resolves the vulnerability.
Is @anthropic-ai/claude-code affected?
Yes, @anthropic-ai/claude-code is affected in releases prior to 2.1.53. Upgrade your dependency to 2.1.53 or later.
What CWE is associated with CVE-2026-33068?
CVE-2026-33068 is classified under CWE-807: Reliance on Untrusted Inputs in a Security Decision (The product uses a protection mechanism that relies on the existence or values of an input, but the input can be modified by an untrusted actor in a way that bypasses the protection mechanism.).
How can I check npm dependencies for CVE-2026-33068?
Run the dependency tree command: 'npm ls @anthropic-ai/claude-code'. This command lists direct and transitive dependencies in your build configuration so you can verify if a vulnerable version is resolved.
Machine-readable OSV JSON OSV v1.5 Schema
Normalized schema output
View Raw JSON 
{
  "schema_version" : "1.4.0",
  "id" : "GHSA-mmgp-wc2j-qcv7",
  "modified" : "2026-03-20T21:24:19Z",
  "published" : "2026-03-19T12:42:09Z",
  "aliases" : [ "CVE-2026-33068" ],
  "summary" : "Claude Code has a Workspace Trust Dialog Bypass via Repo-Controlled Settings File",
  "details" : "Claude Code resolved the permission mode from settings files, including the repo-controlled `.claude/settings.json`, before determining whether to display the workspace trust confirmation dialog. A malicious repository could set `permissions.defaultMode` to `bypassPermissions` in its committed `.claude/settings.json`, causing the trust dialog to be silently skipped on first open. This allowed a user to be placed into a permissive mode without seeing the trust confirmation prompt, making it easier for an attacker-controlled repository to gain tool execution without explicit user consent.\n\nUsers on standard Claude Code auto-update have received this fix already. Users performing manual updates are advised to update to the latest version.\n\nThank you to hackerone.com/cantina_xyz for reporting this issue.",
  "severity" : [ {
    "type" : "CVSS_V4",
    "score" : "CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N"
  } ],
  "affected" : [ {
    "package" : {
      "ecosystem" : "npm",
      "name" : "@anthropic-ai/claude-code"
    },
    "ranges" : [ {
      "type" : "ECOSYSTEM",
      "events" : [ {
        "introduced" : "0"
      }, {
        "fixed" : "2.1.53"
      } ]
    } ]
  } ],
  "references" : [ {
    "type" : "WEB",
    "url" : "https://github.com/anthropics/claude-code/security/advisories/GHSA-mmgp-wc2j-qcv7"
  }, {
    "type" : "ADVISORY",
    "url" : "https://nvd.nist.gov/vuln/detail/CVE-2026-33068"
  }, {
    "type" : "PACKAGE",
    "url" : "https://github.com/anthropics/claude-code"
  } ],
  "database_specific" : {
    "cwe_ids" : [ "CWE-807" ],
    "severity" : "HIGH",
    "github_reviewed" : true,
    "github_reviewed_at" : "2026-03-19T12:42:09Z",
    "nvd_published_at" : "2026-03-20T09:16:15Z"
  }
}