Executive brief
FoundDream miniclawd is a lightweight AI assistant that supports multiple large language models and automated tools. A security flaw in its command execution tool allows remote users to run unauthorized operating system commands on the host server. This could lead to full system takeover, data theft, or disruption of the assistant's services.
Technical details
An OS command injection vulnerability exists in FoundDream miniclawd up to commit 2d65665. The vulnerability is located in the ExecTool.execute function within /src/tools/exec.ts, which utilizes the Node.js spawn() method with the 'shell: true' option enabled. Because the application fails to validate or sanitize the 'command' parameter before passing it to the shell, a remote attacker can inject shell metacharacters (such as semicolons or pipes) to execute arbitrary commands. This can be triggered by sending crafted prompts to the AI assistant that cause it to invoke the execution tool with malicious input. As of the advisory date, no official patch has been released.
Affected products
- FoundDream miniclawd up to 2d65665046e2222eeea76cafc8570ed546a8c125
Timeline
- 2026-04-23: disclosed: Issue reported to the project maintainers via GitHub
- 2026-05-25: advisory: Vulnerability published by VulDB and NVD