atlas-marketing-studio

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime LLM ingestion of outsider-authored free text occurs via the user-controlled request JSON fields like topic and productIdea in the API route /api/generate-ad-skit (and similarly topic in drama ad generation), which are fed into generateSkitScript/script generation prompts.

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

• Direct money access detected (high risk: 1.00). The skill includes explicit payment integration and credit-management code. It instantiates the Stripe SDK with process.env.STRIPE_SECRET_KEY and calls stripe.checkout.sessions.create(...) to create payment checkout sessions (direct use of a payment gateway API). It also contains server-side functions that deduct and add user credits (deductCredits, redeem code handling) and pricing/credit-pack configuration tied to payments. These are specific financial execution capabilities (creating payment sessions and updating balances), so this qualifies as Direct Financial Execution.