GHSA-f934-5rqf-xx47: Arbitrary Workspace File Read via Path Restriction Bypass in OpenClaw
Vulnerability ID: GHSA-F934-5RQF-XX47
CVSS Score: 4.3
Published: 2026-04-17
OpenClaw versions prior to 2026.4.14 contain an improper path limitation vulnerability in the QMD memory management component. The memory_get tool allows authenticated actors to bypass intended intra-workspace access controls and read arbitrary Markdown files stored within the application workspace.
TL;DR
A path restriction bypass in OpenClaw's QMD memory component allows authenticated attackers to read arbitrary Markdown files within the application workspace via the memory_get tool. The issue is patched in version 2026.4.14 by introducing strict contextual validation for file paths.
⚠️ Exploit Status: POC
Technical Details
- Vulnerability Class: Path Restriction Bypass (CWE-22)
- CVSS v3.1: 4.3 Medium
- Attack Vector: Network (Authenticated)
- Impact: Workspace Information Disclosure
- Exploit Status: PoC Available
- Affected Component: QmdMemoryManager
Affected Systems
- OpenClaw QMD Memory Manager
- openclaw npm package
-
openclaw: < 2026.4.14 (Fixed in:
2026.4.14)
Code Analysis
Commit: 37d5971
Fix QMD memory_get restricting reads to canonical or indexed memory paths
Mitigation Strategies
- Update the openclaw package to version 2026.4.14 or later
- Isolate sensitive documentation and secrets from the OpenClaw workspace boundary
- Implement application-level monitoring for anomalous memory_get tool invocations
Remediation Steps:
- Upgrade the openclaw dependency via npm to version 2026.4.14
- Audit all existing workspace directories to identify and remove sensitive data stored in Markdown format
- Verify that the application successfully rejects requests for files outside the memory/ directory
- Deploy SIEM alerting for any 'path required' errors originating from the QmdMemoryManager, which may indicate exploit attempts
References
Read the full report for GHSA-F934-5RQF-XX47 on our website for more details including interactive diagrams and full exploit analysis.
Top comments (0)