lulo

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches serialized transactions at runtime from https://api.lulo.fi (and uses https://blink.lulo.fi for Blink actions), and those responses are deserialized, signed, and submitted—meaning remote content directly determines on-chain instructions the agent executes and is a required runtime dependency.

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

• Direct money access detected (high risk: 1.00). The skill explicitly provides on-chain financial execution capabilities. It defines API endpoints to "generate deposit transaction" (POST /v1/generate.transactions.deposit) and "generate withdrawal transaction" (POST /v1/generate.transactions.withdraw) that return serialized transactions, and the documentation includes concrete TypeScript and Python examples that deserialize, sign with a private key/wallet, and send those transactions to the Solana network. It also lists supported tokens (USDC, USDT, SOL, etc.), shows blink/action endpoints that execute lending actions, and includes a SolanaAgentKit example (agent.methods.lendAssets) to perform lending. These are specific crypto/blockchain transaction tools to move funds, so this qualifies as Direct Financial Execution.