plugin‑shell’s run_command function wrapped every agent‑supplied command with 'sh -c' or 'cmd /C' and forwarded the full argument string to the shell parser, allowing shell metacharacters to be interpreted as shell syntax. This enabled attackers to inject arbitrary commands into the shell, leading to remote code execution and compromising the host’s confidentiality, integrity, and availability. The weakness corresponds to command injection (CWE‑77 and CWE‑78).

Dragonmonk111 JunoClaw, the agentic AI platform built on the Juno Network, is affected in all releases prior to v0.x.y-security-1. The plugin‑shell component that processes agent commands is the vulnerable part.

The CVSS score of 8.4 indicates high severity. Although an EPSS score is not available, the vulnerability is not listed in the CISA KEV catalog, suggesting no documented exploitation to date. The likely attack vector is a remote attacker sending crafted agent‑supplied commands through the platform’s API or command interface, which will be interpreted by the shell wrapper and executed with the privileges of the service.