vgpu-agent-flow
Warn
Audited by Socket on Oct 8, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS. The workflow is internally coherent for multi-agent repo orchestration, but its trust model is not proportionate: an unverifiable external CLI sits in the critical path and receives project OIDC credentials plus broad repository context. No confirmed malware or explicit exfiltration is shown, but the install/provenance gap and credential forwarding make this a high-risk skill.
Confidence: 86%Severity: 88%
Audit Metadata