gpt-image
Warn
Audited by Socket on Jun 19, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The stated purpose is image generation, but the skill delegates execution, authentication, and data flow to inference.sh via a separately installed CLI skill instead of using OpenAI directly. This is not clearly malicious, and the publisher relationship appears coherent, but the intermediary data path, credential forwarding, transitive skill installation, and broad `belt *` permission make the footprint larger than a narrowly scoped image skill.
Confidence: 84%Severity: 64%
Audit Metadata