Executive brief
zevorn rt-claw is an AI assistant framework. A security flaw in its communication system allows unauthorized remote users to bypass security restrictions and execute commands on the host system. This could lead to a complete takeover of the device running the software, potentially exposing sensitive data or allowing the device to be used in further attacks.
Technical details
An incorrect authorization vulnerability exists in the `claw_tool_invoke` function within `claw/services/swarm/swarm.c` of zevorn rt-claw. While the software includes a `CLAW_TOOL_LOCAL_ONLY` flag intended to prevent remote execution of sensitive tools, the RPC receiver path fails to validate this flag upon receiving inbound UDP requests. An unauthenticated remote attacker can send a crafted UDP swarm RPC request to the public listener (typically port 5300) to invoke the `run_script` tool. On Linux systems, this results in arbitrary Python code execution via `popen()`. As of the advisory date, no patch has been released by the vendor.
Affected products
- zevorn rt-claw up to 0.2.0
Timeline
- 2026-07-19: disclosed: Public disclosure of the vulnerability and exploit details.
- 2026-07-19: advisory: NVD and VulDB published the advisory.