fittings-devicerecall-search

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

• Third-party content exposure detected (low risk: 0.10). At runtime the agent workflow calls GET https://fittings.sh/v1/devicerecall/search (FDA recall data) using user-provided query parameters like query/firm, but the only outsider-authored free text ingested is the caller-supplied structured fields used for searching—not content generated by an outsider that is fetched/ingested as free text without selecting a specific result.

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

• Direct money access detected (high risk: 1.00). The skill explicitly includes a payment flow: the endpoint can return a signed quote (402), the client signs an authorization, and a facilitator settles the payment. The documentation shows a concrete client (call.mjs) that uses an on-chain wallet (export FITTINGS_PRIVATE_KEY=0x...), requires USDC on Base, and references EVM tooling (@x402/evm). These are specific financial-execution mechanisms (signing/authorizing payments via a wallet), not mere generic HTTP or browsing capabilities.