integrate-reflect

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's Fresh Context Policy explicitly instructs the agent to fetch live documentation at https://docs.reflect.money at runtime as the "source of truth," meaning remote content from that URL can directly alter agent instructions and is relied on for correct behavior.

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

• Direct money access detected (high risk: 1.00). The skill is explicitly a crypto-financial integration for minting/redeeming yield-bearing stablecoins, whitelabeling branded stablecoins, restaking, and claiming fees on Solana. It exposes concrete SDK methods and REST endpoints that create/send transactions moving value: e.g., UsdcPlusStablecoin.mint(), .redeem(), Proxy.deposit()/withdraw(), /stablecoin/mint, /stablecoin/burn, /integration/mint, /integration/flow/mint, rlp.restake()/withdraw(), and claimFees()/integration/claim. These are not generic callers — they are purpose-built blockchain payment/asset operations (minting, burning, restaking, transferring authority, and fee claims). Under the "Specificity vs Genericity" rule, this is specifically designed to move financial assets on-chain, so it grants Direct Financial Execution Authority.