awesome-tools

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

• Potentially malicious external URL detected (medium risk: 0.30). The skill instructs the agent to read LinkDR's SKILL.md and adopt its current services guidance, so this is an inbound instruction dependency from a separate, named third-party service with no strong first-party provenance, making it common/reputable unofficial and medium; the awesome.tools endpoints are first-party and therefore not counted.

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

• Direct money access detected (medium risk: 0.30). Decisive operation: the skill offers paid listing purchases (Featured Launch at 49 USD and Featured + Newsletter Sponsor at 149 USD). Through the MCP submit_tool, the agent can return a Stripe Checkout URL when called with confirmPaymentSpend=true, so it initiates a purchase instruction for downstream payment. This is a direct operation (a purchase), not an indirect one. The agent-supplied confirmPaymentSpend flag and the "requester authorized" instruction are agent-followed controls and do not count as qualifying access control. However, money moves only when the requester completes payment on the Stripe Checkout page, a payment-provider step that requires the payer's own payment entry. That is strong external approval, so the operation falls under Medium rather than High. The free backlink path and the no-payment API preview are not financial operations. Stripe checkout is the only route that moves money.