computer-use
Warn
Audited by Socket on Aug 24, 2026
1 alert found:
SecuritySecurityagents/openai.yaml
MEDIUMSecurityMEDIUM
agents/openai.yaml
No explicit malware payload is shown, but the configuration is a deliberately high-permission, safety-bypassing agent policy that enables unrestricted local Mac UI automation and explicitly includes sensitive actions (logins, installs/deletions, financial operations) without confirmation or safety gates. This materially increases the risk of abuse via prompt injection, social engineering, or unintended execution.
Confidence: 78%Severity: 87%
Audit Metadata