code-review
Warn
Audited by Socket on May 15, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: mostly coherent as a code-review skill, with no direct credential harvesting or exfiltration, but it has two notable risks disproportionate to a pure reviewer: it loads custom guidance from the untrusted PR branch and it encourages installation of another skill, creating a transitive trust chain. Same-org OpenHands references reduce maliciousness concerns, but prompt-influence and trust-extension keep the overall risk at medium.
Confidence: 89%Severity: 58%
Audit Metadata