PT-2026-50970 · Jovancoding+2 · Network-Ai
CVE-2026-54051
·
Published
2026-06-19
·
Updated
2026-07-20
CVSS v3.1
9.9
Critical
| Vector | AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H |
Name of the Vulnerable Software and Affected Versions
Network-AI versions prior to 5.9.1
Description
An issue exists in the agent sandbox where shell commands are gated behind an allowlist using the
SandboxPolicy.isCommandAllowed function. The allowlist uses glob-matching on the entire command string, but the ShellExecutor executes these strings via /bin/sh -c. This discrepancy allows an attacker to bypass the restriction; for example, a wildcard entry like git * would also match git status; <command>, leading to arbitrary command execution as the orchestrator process. This effectively defeats the primary control intended to contain a compromised agent.Recommendations
Upgrade to network-ai version 5.9.1 or later.
As a defense in depth measure, avoid using broad wildcard allowlist entries such as
node * or npm * as these can allow direct code execution by design.Exploit
Fix
OS Command Injection
Found an issue in the description? Have something to add? Feel free to write us 👾
Weakness Enumeration
Related Identifiers
Affected Products
Network-Ai