GHSA-39MP-545Q-W789: Improper Authorization in OpenClaw /send Command
Vulnerability ID: GHSA-39MP-545Q-W789
CVSS Score: 5.4
Published: 2026-03-30
An authorization bypass vulnerability in the OpenClaw personal AI assistant ecosystem allows non-owner users with generic command permissions to persistently modify session-specific message delivery policies. By issuing the /send command, attackers can silence the agent or force unintended message delivery.
TL;DR
OpenClaw versions prior to 2026.3.27 fail to verify session ownership when processing the /send command. This improper authorization allows any command-authorized user in a shared session to alter the agent's delivery policy, leading to localized service disruption and configuration overrides.
Technical Details
- CWE ID: CWE-285 (Improper Authorization)
- Attack Vector: Network
- CVSS v3.1 Score: 5.4 (Medium)
- Impact: Low Integrity, Low Availability
- Exploit Status: None / Theoretical
- CISA KEV: Not Listed
Affected Systems
- OpenClaw personal AI assistant ecosystem
- OpenClaw gateway-chat module
-
openclaw: < 2026.3.27 (Fixed in:
2026.3.27)
Mitigation Strategies
- Upgrade the OpenClaw npm package to version 2026.3.27 or newer.
- Restrict generic command access by auditing the
commands.allowFromconfiguration list. - Ensure all custom command handlers modifying session state enforce
ownerOnlyprivilege checks.
Remediation Steps:
- Identify the current version of the OpenClaw package running in the environment.
- Execute
npm install openclaw@latestto pull the patched release. - Review the
config.jsonfile and minimize the number of users listed incommands.allowFrom. - Restart the OpenClaw service to apply the updated code and configuration.
References
- GitHub Advisory: GHSA-39MP-545Q-W789
- OpenClaw Repository
- Snyk Vulnerability Database
- SOOS Research Vulnerability Entry
- OpenClaw CVE Tracking
Read the full report for GHSA-39MP-545Q-W789 on our website for more details including interactive diagrams and full exploit analysis.
Top comments (0)