The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process potentially untrusted data from user requests, personal profiles, and recall results to inform its search and recommendation logic. This creates a surface where embedded instructions in the source data could potentially influence the agent's behavior.
Ingestion points: The skill reads user-described needs, profile data (such as household information and saved limits), and recall results (as described in SKILL.md).
Boundary markers: The instructions do not define specific delimiters or "ignore" directives to isolate the agent from instructions that might be contained within the customer's phrasing or profile data.
Capability inventory: The agent uses tools including present_products, present_comparison, get_product_details, and web_search to execute actions based on the ingested data.
Sanitization: The skill includes an instruction to "Word the query in the catalog's vocabulary and leave the customer's phrasing behind," which functions as a form of input normalization, though no further sanitization or validation of the external content is specified.

search-discovery