Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only image workflow skill with a capability inconsistency, but no evidence of hidden code, persistence, exfiltration, or destructive behavior.
Before installing, treat the custom-model and paid-feature claims as unreliable until the publisher clarifies them. The skill otherwise appears to be a plain Markdown guide for image-related agent work; approve file writes or shell commands only when they match the specific image task you requested.
The manifest description states the skill supports '自定义模型微调' as an advanced capability. Later documentation says '目前图像技能不支持自定义模型', which directly contradicts that earlier claim rather than merely omitting detail.
The feature matrix marks '自定义模型微调' as supported for the paid tier. The FAQ later states the image skill does not support custom models, which is an active contradiction about the skill's intended behavior.
No suspicious patterns detected.
