ultra-review
Warn
Audited by Socket on Aug 3, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: the core review purpose is legitimate and mostly aligned with read-only code inspection, but the instruction to load every relevant available skill creates a meaningful transitive-trust expansion. External research requirements also introduce indirect prompt-injection risk in a tool-rich harness. No direct malware, credential harvesting, installer abuse, or exfiltration path is present in the supplied skill text.
Confidence: 87%Severity: 56%
Audit Metadata