| Title | SkyworkAI DeepResearchAgent main OS Command Injection |
|---|
| Description | Multiple critical Remote Code Execution (RCE) vulnerabilities exist in the DeepResearchAgent project's tool loading system. These vulnerabilities allow arbitrary code execution through insufficient input validation in the `Tool.from_code()`, `Tool.from_dict()`, and `ToolCollection.from_mcp()` methods. When loading tools from external sources, user-controlled input is directly passed to Python's `exec()` function or system command execution without any sanitization or validation, enabling attackers to execute arbitrary code with the privileges of the application process. |
|---|
| Source | ⚠️ https://github.com/bayuncao-bit/vul-36 |
|---|
| User | bayuncao (UID 50143) |
|---|
| Submission | 07/23/2025 08:57 (9 months ago) |
|---|
| Moderation | 08/06/2025 12:33 (14 days later) |
|---|
| Status | Accepted |
|---|
| VulDB entry | 319026 [SkyworkAI DeepResearchAgent up to 08eb7f8eb9505d0094d75bb97ff7dacc3fa3bbf2 src/tools/tools.py from_code/from_dict/from_mcp os command injection] |
|---|
| Points | 20 |
|---|