setting-up-agent-relay-flows

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

• Third-party content exposure detected (low risk: 0.10). The skill instructs the agent to download flow source code from an external URL (source.rawUrl) from the flow catalog and deploy it.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill instructs the execution of npx --yes agent-relay@13 cloud connect anthropic --api-url 'https://agentrelay.com/cloud', which fetches and runs an external package at runtime, presenting a weak/unknown/suspicious operational dependency since package execution provenance is not strongly authenticated.

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

• Direct money access detected (medium risk: 0.30). The documented workflow sets up agent relay flows and integrations via API calls (such as activating flows through POST /api/v1/flows/deploy and connecting tools). Although it mentions funding promotional runs (SKILL.md:83, 138-142), it does not perform direct financial transfers, charges, purchases, trades, or settlement authorization; nor does it submit bills, invoices, or payroll. Therefore, the operations are indirect service/subscription usage funding rather than direct financial transactions.