payermax-integration-assistant

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt mandates generating a configuration file that includes sensitive credentials (appId, merchantNo, merchant-private-key, merchant-public-key, payermax-public-key, etc.) and even instructs generating/validating and embedding Base64 RSA key material into output, which requires the LLM to produce secret values verbatim.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly mandates fetching remote integration docs at runtime (e.g., https://docs.payermax.com/.../*.md) and may load the Drop-In JS SDK in generated frontend code (https://cdn.payermax.com/dropin/js/pmdropin.min.js), both of which are required runtime dependencies that directly control implementation instructions and (in the case of the CDN) execute remote code.

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

• Direct money access detected (high risk: 1.00). The skill is explicitly a payment integration assistant for PayerMax. It defines API endpoints and workflows that create and manage real payments and refunds (e.g., orderAndPay, createPaybylink, subscriptionCreate, /refund, /refundQuery, /orderQuery), requires merchant credentials and RSA signing keys, and mandates generating production-ready code to create payments, trigger refunds, and perform subscription debits. It also requires connectivity tests that send signed requests to sandbox endpoints. These are specific tools/actions whose primary purpose is to move money (create charges, process refunds, initiate recurring debits), so it grants direct financial execution authority.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)