sumup-best-practices

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 to "always prefer the latest SumUp docs" and specifies an LLM entrypoint URL (https://developer.sumup.com/llms.txt) and docs root (https://developer.sumup.com/), indicating the agent will fetch remote markdown/prompt content at runtime that can directly control prompts.

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

• Direct money access detected (high risk: 1.00). The skill is explicitly focused on integrating SumUp (a payment gateway) and references specific payment APIs and SDKs used to accept and process payments: Hosted Checkout, Card Widget, Checkouts API (server-to-server), Customers + tokenization (save-and-charge), Cloud API, and Terminal/Reader SDKs. It includes operational directives that enable transaction creation and lifecycle handling (webhooks, unique transaction references, 3DS, charging saved cards) and directs hand-off to implementation skills for end-to-end payment flows. This is a purpose-built payment integration skill (not a generic tool), so it grants direct financial execution capability.