| Title | 0x4m4 hexstrike-ai v1.0 RCE |
|---|
| Description | The HexStrike AI server exposes 90+ tool execution endpoints under /api/tools/* (e.g., /api/tools/nmap, /api/tools/hydra, /api/tools/sqlmap, /api/tools/msfvenom, /api/tools/metasploit). Every endpoint follows the same pattern: user-supplied parameters (especially additional_args, but also target, username, password, scan_type, payload, etc.) are concatenated directly into a shell command string using Python f-strings, then executed via subprocess.Popen(..., shell=True). No input sanitization, escaping, or allowlisting is performed. The server requires no authentication and binds to 0.0.0.0.
An attacker can inject arbitrary shell commands by appending shell metacharacters (;, |, &&, $(), backticks) in any of the user-controllable fields. Each of the 90+ endpoints is independently vulnerable. |
|---|
| Source | ⚠️ https://github.com/0x4m4/hexstrike-ai/issues/224 |
|---|
| User | ez-lbz (UID 87033) |
|---|
| Submission | 08/04/2026 14:34 (1 month ago) |
|---|
| Moderation | 09/13/2026 07:08 (1 month later) |
|---|
| Status | Accepted |
|---|
| VulDB entry | 403223 [0x4m4 HexStrike AI up to d689933ff579d839c676c82b231f8e98326c5f04 API Tools Endpoint hexstrike_server.py subprocess.Popen os command injection] |
|---|
| Points | 20 |
|---|