converging-and-polishing
Warn
Audited by Socket on Jul 7, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: the core purpose is coherent, but the skill meaningfully expands agent autonomy and trust boundaries. The main risks are transitive skill loading, endless unattended operation, and external transmission of project content to reviewer CLIs; install provenance for those CLIs appears official rather than attacker-controlled.
Confidence: 87%Severity: 68%
Audit Metadata