improve-skills
Warn
Audited by Socket on Jul 3, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The skill is largely coherent with an internal AllenAI skill-improvement workflow: repo cloning, issue/PR creation, log inspection, and skill edits all match the stated purpose. Main concerns are broad action scope, credential export to internal tooling, and moderate supply-chain trust from unpinned tool installs and partially unverifiable internal CLI provenance; however, there is no strong sign of credential harvesting, unrelated access, or deceptive exfiltration.
Confidence: 80%Severity: 56%
Audit Metadata