agentic-loop
Warn
Audited by Socket on Jul 18, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: the skill’s capabilities mostly match its stated memory-server purpose, but it expands trust to a third-party MCP server fetched and executed via unpinned `npx`. Data flows are broadly proportionate to persistent-memory functionality, yet the server receives detailed task history and can influence future agent behavior, creating medium supply-chain and autonomy risk.
Confidence: 82%Severity: 56%
Audit Metadata