Openclaw cli: OpenClaw exec approvals: safeBins could bypass stdin-only constraints via shell expansion (CVE-2026-28463)
Description
OpenClaw CLI versions before 2026.2.14 have a vulnerability in the exec-approvals allowlist feature where safe binaries intended to be stdin-only can bypass restrictions via shell expansion. This allows authorized but untrusted users to cause the gateway or node process to read arbitrary local files when host execution is enabled in allowlist mode. The issue arises because the allowlist validation checks tokens before shell expansion, but execution uses a real shell that expands globs and environment variables. A fix is available in version 2026.2.14 that forces argv tokens to be treated as literal text, preventing this bypass.
CVSS v3.1
Score 5.7medium
Affected software
Run on your own infrastructure? Check whether these packages are installed with threat-finder — our free open-source scanner.
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
OpenClaw's exec-approvals allowlist supports a set of safe binaries intended to accept only stdin input without positional file arguments when running with host execution enabled (`tools.exec.host=gateway|node`) and security set to allowlist. The vulnerability occurs because the allowlist validation checks the command arguments before shell expansion, but the actual execution uses `sh -c`, which expands globs (e.g., `*`) and environment variables (e.g., `$HOME`). This discrepancy allows safe binaries like `head`, `tail`, or `grep` to read arbitrary local files by exploiting shell expansions, bypassing the intended stdin-only constraint. This issue is configuration-dependent and not triggered by default settings where `tools.exec.host` is `sandbox`. The vulnerability affects openclaw-cli versions from 2026.1.29 up to but not including 2026.2.14. The fix, planned for 2026.2.14, involves forcing argv tokens to be single-quoted at execution time to prevent shell expansion.
Potential Impact
An authorized but untrusted caller or prompt injection can exploit this vulnerability to cause the gateway or node process to disclose files readable by that process. This leads to unauthorized local file disclosure when host execution is enabled in allowlist mode. The impact is limited to configurations where host execution is enabled and the allowlist mode is active. The CVSS score is 5.7 (medium severity) with high confidentiality impact but no integrity or availability impact.
Mitigation Recommendations
A patch is available in openclaw-cli version 2026.2.14 and later that fixes the issue by forcing argv tokens to be treated as literal text during execution, preventing shell globbing and environment variable expansion. Users should upgrade to version 2026.2.14 or later once released. Until then, avoid enabling host execution with allowlist mode or restrict usage to trusted callers. The issue is not exercised by default settings, so default configurations are not vulnerable.
Technical Details
- Gcve Source
- db.gcve.eu
- Osv Id
- BREW-openclaw-cli-CVE-2026-28463
- Osv Schema Version
- 1.7.3
- Ecosystems
- ["Homebrew"]
- Cvss Version
- 3.1
Threat ID: 6ac245c712601ec6a3167f06
Added to database: 10/04/2026, 12:25:43 UTC
Last enriched: 10/04/2026, 13:09:49 UTC
Last updated: 10/04/2026, 16:08:37 UTC
Views: 2
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.