usda-food-search

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). This set includes a direct downloadable script (https://fittings.sh/skill/call.mjs) that the README explicitly tells users to curl and run (node call.mjs), which is a high-risk pattern for distributing malware; the other listed URLs are API endpoints and are low risk by themselves.

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

• Third-party content exposure detected (low risk: 0.10). In the required runtime workflow, user-supplied query parameters (e.g., query, dataType, limit) are sent as part of the request to https://fittings.sh/v1/usda/food-search, but there is no step where an outsider-authored free-text source (like email/chat/inbox/tickets/issues) is ingested without first selecting the specific USDA lookup item.

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

• Potentially malicious external URL detected (high risk: 0.90). The SKILL instructs fetching and running remote code (curl -sO https://fittings.sh/skill/call.mjs) which, if executed, would run code obtained at runtime from that URL.

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

• Direct money access detected (high risk: 1.00). The skill includes an explicit payment flow that requires a private wallet key and signing authorizations (uses USDC on Base, FITTINGS_PRIVATE_KEY, node call.mjs). It references EVM libraries (@x402/evm) and describes signing a quote and having a facilitator settle the payment — i.e., explicit crypto/wallet signing capability that can execute payments. This is a specific crypto/blockchain payment integration, so it grants direct financial execution authority.