The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements an AI-assisted data enrichment pattern in references/bulk-catalog-enrichment.md that is vulnerable to indirect prompt injection (Category 8).
Ingestion points: The generateDescription and enrichSEOMetadata functions ingest untrusted data from product names, categories, and attributes stored in the commercetools catalog.
Boundary markers: Absent. The product data is directly interpolated into the prompt sent to the LLM without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill uses the batchUpdateProducts utility to write AI-generated content (descriptions and SEO metadata) back to the live commercetools catalog using the setDescription, setMetaTitle, and setMetaDescription update actions.
Sanitization: Absent. There is no evidence of escaping, validation, or filtering of the ingested product content before it is sent to the LLM provider.

commercetools-data