clearhaus

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill calls the Membrane CLI at runtime (e.g., membrane connection ensure "https://www.clearhaus.com/") and Membrane's responses can include clientAction.agentInstructions (and clientAction.uiUrl) which are fetched remotely and can directly instruct the AI agent, so the https://www.clearhaus.com/ URL (and the Membrane-hosted connection responses) represent a runtime external dependency that can control agent prompts.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill is explicitly a Clearhaus payment-gateway integration and exposes payment-specific actions (e.g., "Charge", "Refund") and the ability to run actions or proxy API requests that can create transactions. It uses Membrane to handle auth/credentials so an agent can invoke those payment APIs directly (membrane action run, membrane request), enabling creation or reversal of payments. This is specifically designed to move money, not a generic tool, so it grants direct financial execution capability.