gpt-image-2
Warn
Audited by Socket on May 4, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS: the image-editing purpose is plausible, but the default behavior routes OpenAI credentials and image data through an unrelated third-party host and encourages shell-sourced secret access. Combined with a mutable clone-and-run install path from a personal repo, the skill's trust and data-flow footprint are not proportionate to a normal OpenAI image client.
Confidence: 92%Severity: 90%
Audit Metadata