lending

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs running network-backed discovery/status commands (e.g., "npx starkfi@latest lend-pools" and "npx starkfi@latest lend-status") which fetch live, public protocol/pool data that the agent must read and use (health-factor and pool info) to decide lending actions, so untrusted third-party content can materially influence behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill repeatedly requires running "npx starkfi@latest", which at runtime fetches and executes remote package code from the npm registry (e.g. https://registry.npmjs.org/starkfi), so this external fetch executes remote code and is a required dependency.

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

• Direct money access detected (high risk: 1.00). The skill is explicitly designed to perform on-chain DeFi lending operations on Starknet. It contains concrete transactional commands (npx starkfi lend-supply, lend-borrow, lend-repay, lend-withdraw, lend-close) that move crypto assets, requires an active wallet session and gas/token balances, and references tx-status verification. This is a specific crypto/blockchain execution capability (signing/sending transactions and managing loans), so it grants direct financial execution authority.