orchestrator
Warn
Audited by Socket on Mar 15, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS: the skill's purpose matches orchestration, but it enables high-autonomy execution through external CLI subagents with `--approval-mode=yolo` and feeds untrusted agent-generated content back into the control loop. No clear credential theft or exfiltration is shown, but execution trust and indirect prompt-injection risk are material.
Confidence: 83%Severity: 74%
Audit Metadata