The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes catalog data which is considered untrusted content from external merchant sources.
Ingestion points: Catalog tools such as search_products and get_product in references/tools.md process external store data.
Boundary markers: The implementation uses untrustedContentHint: true and readOnlyHint annotations to help the agent distinguish content from instructions.
Capability inventory: The skill provides tools for cart management (add_to_cart, remove_from_cart) and site navigation (navigate) as defined in references/tools.md.
Sanitization: Upstream error messages are truncated before being presented to the agent to minimize potential context flooding.
[EXTERNAL_DOWNLOADS]: The skill instructions recommend installing the @commercengine/ai package and its peer dependencies from the npm registry. These are official vendor packages designed for storefront integration.

ce-ai