agentation
Warn
Audited by Socket on Mar 18, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The core capability is coherent with a UI annotation skill, and most data flow is local-first, but the footprint is broader than a simple UI helper: it installs/executes npm packages via unpinned `npx`, modifies multiple agent configs, supports transitive skill installation, enables autonomous edit loops, and can forward annotation data to arbitrary webhooks. This looks more like high-risk automation tooling than outright malware.
Confidence: 82%Severity: 66%
Audit Metadata