sales-checkoutpage

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required workflow includes handling Checkout Page conversion webhook deliveries (and MCP tool calls) whose JSON data can contain free-form customer/form fields (e.g., names, emails, attendee responses) that are outsider-authored, and these values are typically ingested into the agent/handler context at runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent at runtime to add and use a hosted MCP (claude mcp add ...) at https://mcp.checkoutpage.com, which is an external agentic tool endpoint that the assistant will call to execute actions on behalf of the agent.

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

• Direct money access detected (high risk: 1.00). This skill is explicitly tied to a Stripe-based payments platform and documents REST/MCP operations and API keys/OAuth for Stripe. It references payment-specific endpoints (GET /v1/payments, GET /v1/customers), Stripe object ids, and instructs using the native MCP to create paid checkout pages (create_checkout_page/create_coupon) end-to-end from a prompt. Those are specific payment-gateway integrations and agent-capable actions that can create and manage revenue-producing checkout flows — i.e., direct financial execution via Stripe.