pentagi

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). High abuse potential: autonomous agents can craft and run exploits (metasploit, sqlmap, payloads), persist successful techniques/credentials in a knowledge graph, and be triggered via API—features that enable unauthorized data exfiltration, remote code execution, and repeated compromise if misused.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The instructions explicitly run git clone https://github.com/vxcontrol/pentagi.git as part of deployment and then run the repository's code via docker compose up -d, so this external GitHub URL fetches required remote code that will be executed at runtime.