israeli-marketplace-seller
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data during its core operations, creating a potential surface for indirect prompt injection.
- Ingestion points: The skill uses the
WebFetchtool to monitor competitor prices on public marketplaces (Zap, KSP) and aggregates customer inquiries and reviews from Meta (Facebook, Instagram) and Zap dashboards. - Capability inventory: The agent has the ability to fetch web content, write to local JSON files (
price-history.json), and draft Hebrew responses for customer communication. - Boundary markers: The instructions do not define explicit delimiters or boundary markers to separate untrusted external content from system instructions when processing price data or message drafts.
- Sanitization: There are no instructions for sanitizing or filtering instructions that might be embedded in product descriptions, reviews, or customer inquiries before the agent processes them.
Audit Metadata