Bankr Dev - Client Patterns

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

• Third-party content exposure detected (high risk: 0.80). The client explicitly fetches and ingests responses from an external/public API (e.g., fetch calls to ${API_URL}/agent/prompt and ${API_URL}/agent/job/${jobId} defaulting to https://api.bankr.bot) and then reads/returns fields like response, statusUpdates, richData (social-card text, chart URLs) and transaction messages — all of which are untrusted third-party content that the agent will interpret at runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill sends user prompts at runtime to the Bankr API (default URL https://api.bankr.bot, e.g. POST to https://api.bankr.bot/agent/prompt) and depends on the remote service's responses to drive agent behavior and job execution, so the fetched content directly controls prompts/results 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 is a specialized API client for "Bankr" and explicitly models blockchain/crypto transactions (SwapTransaction, TransferEthTransaction, TransferErc20Transaction, approval/permit metadata, buy_nft, swapCrossChain, etc.). The client exposes functions that submit prompts to the Bankr agent (submitPrompt/execute) and examples show requesting actions like "Buy $50 of ETH on Base" and handling resulting transactions. This is not a generic HTTP helper or browser automation tool — it is explicitly designed to interact with a financial/crypto execution service and to produce/send transaction objects. Therefore it grants direct financial execution capability (crypto/blockchain operations).