The Agent Skills Directory

[SAFE]: The skill implements security best practices for handling large external datasets, including streaming CSV parsing (via csv-parse and fs.createReadStream) to prevent memory exhaustion and strict schema validation (via zod) to ensure data integrity during imports.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted product data from external files (CSV, XML, JSON). While the provided code implements robust schema validation and type coercion for fields like price and handle, the raw text content from product titles or descriptions remains a potential surface for indirect prompt injection if the database records are later consumed by an LLM without proper boundary markers.

catalog-import-export