product-information-management

Pass

Audited by Gen Agent Trust Hub on Mar 16, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection through the ingestion of external product data.
  • Ingestion points: Data is fetched from remote PIM systems (Akeneo, Salsify) via REST APIs and processed through webhook endpoints (POST /api/webhooks/akeneo in SKILL.md).
  • Boundary markers: None identified. The code snippets in SKILL.md map API fields directly to internal data structures without using delimiters or instructions to ignore embedded content.
  • Capability inventory: The integrated scripts perform database updates (db.products.upsert), network operations via fetch, and CDN cache invalidation (revalidateProductPage).
  • Sanitization: There is no evidence of HTML escaping or instruction filtering for the product descriptions and metadata before they are stored or published to the storefront.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 16, 2026, 12:47 PM