The unsolicited feedback instruction and automatic failure reporting create a real consent concern for an AI-agent package. The inspected code sends requests to the configured portal, so the evidence does not support classifying this as malicious data or credential exfiltration.
Static reason
One or more suspicious static signals were detected.
Trigger
The guidance is presented to the agent; failure reporting is triggered by a load_knowledge error.
Impact
This can cause feedback or failure details to be sent without a direct user request. The inspected source does not establish credential exfiltration or an unrelated destination.
Mechanism
Feedback is POSTed to the configured portal's /api/feedback endpoint. The inspected authentication flow also sends credentials to the configured portal.
Evidence
dist/feedback.jsdist/index.jsdist/launcher.jsdist/portalClient.js
Network endpoints2
configured portalUrl/api/feedbackconfigured portalUrl