run-dogfood
Warn
Audited by Socket on Oct 6, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS. The skill's overall purpose is coherent for acceptance testing, but it authorizes broad unattended actions: driving a real signed-in browser session, making product changes, fixing defects, and suppressing interim user interaction until completion. The command examples are not malicious by themselves, and the Kimi installer appears official, but the combination of autonomous product mutations plus reliance on an external browser bridge makes this a high operational-security risk skill rather than a benign low-risk guide.
Confidence: 88%Severity: 76%
Audit Metadata