investigating-eval-results

Warn

Audited by Socket on Oct 21, 2026

1 alert found:

Anomaly
AnomalyLOW
SKILL.md

Mostly coherent dev tooling for local eval investigation, but it is not purely read-only: it drives autonomous edit/rerun loops, processes untrusted trajectory/log content, and instructs use of another skill for self-modification. No clear credential harvesting or attacker-controlled exfiltration is present, so this is better classified as suspicious/medium-risk rather than malicious.

Confidence: 87%Severity: 56%
Audit Metadata
Analyzed At
Oct 21, 2026, 01:36 PM
Package URL
pkg:socket/skills-sh/googlechrome%2Fmodern-web-guidance-src%2Finvestigating-eval-results%2F@606d504c33888b25c157d130633e4a6c9f8ff9acd233b5181fb7befc8f51f564