anthropic
Warn
Audited by Socket on Apr 30, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: The skill is internally coherent as a Membrane connector, and its install source appears publisher-aligned and official enough to avoid a malware judgment. However, it markets Anthropic access while routing auth and API activity through Membrane rather than Anthropic’s native API flow, creating a third-party intermediary for credentials and data. Combined with the unpinned global CLI install and dynamic action creation, this makes the skill medium risk rather than benign.
Confidence: 86%Severity: 64%
Audit Metadata