mp-commerce-checkout
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the official vendor-provided package mercadopago-commerce-agents-checkout==0.1.0. It also references documentation and API endpoints hosted on official Mercado Pago domains (mercadopago.cl, mercadopago.com) and GitHub repositories. These resources are consistent with the skill's purpose and the author's identity.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface where untrusted data (cart contents and products) is processed to generate payment orders.
- Ingestion points: Cart mapping process in Phase 1 and Phase 2 (SKILL.md).
- Boundary markers: The skill explicitly requires a 'handoff boundary' to deliver the payment URL outside of model prompts or memory, preventing the model from acting upon or leaking the URL.
- Capability inventory: The skill uses package managers (uv, poetry, pipenv) to install dependencies and generates a local Python adapter file (Phase 2).
- Sanitization: The skill mandates a 'trusted catalog' lookup to independently verify prices and currency, ensuring that instructions or price overrides embedded in the input cart data are ignored in favor of authoritative server-side data.
Audit Metadata