| Title | assafelovic gpt-researcher 3.4.3 Unauthenticated Remote Code Execution |
|---|
| Description | GPT Researcher v3.4.3 and earlier versions are vulnerable to **unauthenticated Remote Code Execution (RCE)** via the WebSocket `/ws` endpoint. An attacker can send a WebSocket message containing a `mcp_configs` parameter with an arbitrary `command` and `args` array. These values are passed through the application without any validation or sanitization, ultimately reaching `anyio.open_process()` which spawns the attacker-specified process on the server. No authentication is required. The attacker has full control over the executed command, its arguments, and the environment variables of the spawned process. |
|---|
| Source | ⚠️ https://github.com/assafelovic/gpt-researcher/issues/1694 |
|---|
| User | Yu-Bao (UID 96702) |
|---|
| Submission | 03/23/2026 03:31 (15 days ago) |
|---|
| Moderation | 04/05/2026 21:12 (14 days later) |
|---|
| Status | Accepted |
|---|
| VulDB entry | 355419 [assafelovic gpt-researcher up to 3.4.3 ws Endpoint server_utils.py extract_command_data args code injection] |
|---|
| Points | 20 |
|---|