claude-code-wrapper
Warn
Audited by Socket on Oct 7, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The skill is purpose-aligned as a convenience wrapper, but it reads raw Claude credentials from disk and forwards them into a child CLI, with final data destination controlled by ANTHROPIC_BASE_URL. No clear malware or deceptive install behavior is present, but credential handling and indirect endpoint routing create meaningful security risk.
Confidence: 86%Severity: 57%
Audit Metadata