soskills-ecommerce

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly asks for a Stripe publishable key and a Stripe price ID and instructs replacing placeholders and writing those values into .env files and Vercel environment variables, which requires the LLM to accept and output key/ID values verbatim.

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md instructs the agent to perform runtime web research (e.g., “recherche sur Reddit… X… et le web” and extracting info from Amazon/supplier URLs), meaning outsider-authored free text from public webpages/feeds can be ingested into the agent’s LLM context via the research results.

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

• Direct money access detected (high risk: 1.00). The skill explicitly integrates a payment gateway (Stripe). It requests a Stripe publishable key and Stripe price ID, includes @stripe/stripe-js in dependencies, implements a Checkout page that redirects to Stripe Checkout, and the DEPLOYER instructions walk the user through creating Stripe products and swapping test/live keys. These are specific payment gateway actions (creating prices, configuring keys, and initiating Stripe checkout), so the skill provides direct financial execution capability.