secure-checkout

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

• Third-party content exposure detected (high risk: 0.80). The skill's task (evals/payment-tokenization-and-input-validatio/task.md and inputs/product.json) requires fetching and rendering product descriptions from a headless CMS that does not sanitize user-generated HTML (including a tag), so the agent will ingest untrusted third-party content as part of its workflow.

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

• Direct money access detected (high risk: 1.00). The skill explicitly references and shows integration code for payment gateways (Stripe, Stripe Checkout, PayPal) — including client-side Stripe usage (loadStripe, elements, confirmPayment) and paymentIntent handling. Those are specific payment gateway APIs/flows (i.e., moving/confirming payments and handling payment tokens), so this grants direct financial execution capability rather than being a generic tool.