opensea-tool-sdk

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's CLI/SDK fetches tool manifests at runtime from external metadata URLs (e.g., https://my-tool.example.com/.well-known/ai-tool/my-tool.json and https://my-tool.vercel.app/.well-known/ai-tool/my-tool.json), and those manifests directly describe the tool's inputs/outputs and discovery info that agents use to determine how to call or behave, so remote manifest content can control agent prompts/instructions.

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

• Direct money access detected (high risk: 1.00). The skill explicitly exposes on-chain and payment APIs and CLI commands that perform crypto/banking actions: x402 pay-per-call (USDC) and EIP-3009 signing, a pay/paidFetch flow that signs and submits payments, onchain ToolRegistry.registerTool transactions (requires a wallet/private key), wallet providers including custodial services (Fireblocks, Turnkey) and a Bankr HTTP API (createBankrAccount). These are specific crypto/banking/payment capabilities (signing/submitting transfers and interacting with payment/banking providers), so it grants direct financial execution authority.