swiggy-prava-skill

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill requires the agent to obtain and place Prava tokenized card credentials (token, cryptogram, expiry) into the browser checkout flow, which forces the LLM to handle sensitive secret values in action payloads (and risks them being emitted verbatim), even though it forbids printing them in chat.

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

• Potentially malicious external URL detected (high risk: 1.00). The setup instructions include an npx command that will fetch and execute remote code from https://github.com/Prava-Payments/prava-skills at runtime (references/setup.md:55), which is a required installation step for the Prava skill.

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

• Direct money access detected (high risk: 1.00). The skill explicitly orchestrates end-to-end card payments: it instructs creating a Prava session (Prava CLI), tokenizing card credentials, and using those tokenized values as card number/CVV/expiry to submit payment through Swiggy's checkout and payment gateway. These are specific payment gateway/transaction actions (creating payment tokens and submitting card transactions), which constitute direct financial execution authority.