skill-optimizer
Warn
Audited by Socket on Apr 14, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: the skill’s core behavior is mostly consistent with optimization/audit, but it has disproportionate visibility into private session transcripts and explicitly extracts Codex base_instructions, creating privacy and hidden-instruction exposure risk. No clear malware or outbound exfiltration is present in the supplied skill, but the transcript-reading and system-prompt-adjacent behavior make it medium risk.
Confidence: 87%Severity: 66%
Audit Metadata