llm-evals
Warn
Audited by Socket on Nov 16, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
The evaluation workflow is largely consistent with its stated purpose, and the supplied text does not show credential theft or concealed actions. The unpinned instruction to install a separate skill from a third-party repository is a meaningful transitive supply-chain risk; verify that skill independently before installing it.
Confidence: 88%Severity: 72%
Audit Metadata