GHSA-j9pv-rrcj-6pfx: Insecure Environment Inheritance and Information Disclosure in OpenClaw
Vulnerability ID: GHSA-J9PV-RRCJ-6PFX
CVSS Score: 8.6
Published: 2026-04-02
OpenClaw versions prior to v2026.3.31 are vulnerable to information disclosure due to insecure environment inheritance in the SSH-based sandbox backends. The application passes the entire parent process environment, including sensitive AI provider API keys, to child processes.
TL;DR
OpenClaw passes unsanitized environment variables to child processes in its SSH sandbox, exposing AI API keys to local and potentially remote attackers. Upgrading to v2026.3.31 patches this via a strict environment filtering utility.
⚠️ Exploit Status: POC
Technical Details
- CWE ID: CWE-214, CWE-200
- Attack Vector: Local / Remote (Conditional)
- CVSS Score: 8.6
- Exploit Status: Proof of Concept
- Impact: High (Information Disclosure)
- Patch Status: Available (v2026.3.31)
Affected Systems
- OpenClaw ssh-backend component
- OpenClaw openshell extension
-
OpenClaw: < 2026.3.31 (Fixed in:
2026.3.31)
Code Analysis
Commit: cfe1445
Fix: Implement sanitizeEnvVars utility to prevent sensitive environment variable inheritance in subprocesses.
Added sanitize-env-vars.ts and modified ssh-backend.ts to apply filtering.
Mitigation Strategies
- Upgrade to patched software version v2026.3.31
- Harden local SSH client configurations to prevent SendEnv variable forwarding
- Implement principle of least privilege for the OpenClaw service account
Remediation Steps:
- Audit all systems running OpenClaw to identify the current software version.
- Update the application to version v2026.3.31 using the standard package manager or deployment pipeline.
- Inspect
/etc/ssh/ssh_configand~/.ssh/configforSendEnvdirectives. - Remove or restrict
SendEnvparameters to prevent accidental exposure of wildcard API keys. - Rotate any API keys (OpenAI, Anthropic) that were previously utilized by the vulnerable application.
References
Read the full report for GHSA-J9PV-RRCJ-6PFX on our website for more details including interactive diagrams and full exploit analysis.
Top comments (0)