research
Warn
Audited by Socket on Apr 28, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The skill's purpose is coherent, but it combines broad ingestion of untrusted external content with shell and file-write capabilities, creating meaningful indirect prompt-injection risk. Install trust is otherwise moderate and mostly limited to cloning arbitrary third-party repos via official Git/GitHub paths, with no evidence of credential harvesting or covert exfiltration.
Confidence: 86%Severity: 62%
Audit Metadata