skills/nexscope-ai/nexscope-ecommerce-skills/ecommerce-ozon-market-keyword-search/Gen Agent Trust Hub
ecommerce-ozon-market-keyword-search
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and displays marketplace keyword data, including Russian queries, Chinese translations, and product titles, which are sourced from external platforms (Ozon and Wildberries). This content is inherently untrusted and could potentially contain malicious instructions intended to influence the agent's behavior during data processing.
- Ingestion points: Marketplace keyword metrics and metadata are fetched via the Seerfar API and processed in
scripts/ozon_market_keyword_search.pybefore being printed to stdout and saved to local JSON files. - Boundary markers:
SKILL.mdincludes explicit instructions for the agent to 'Present data only' and 'show keyword metrics in a clear table without subjective advice', which serves as a prompt-level boundary to mitigate accidental instruction following. - Capability inventory: The associated script is capable of making network POST requests to the NexScope proxy and writing response data to the local project filesystem.
- Sanitization: The skill does not perform explicit character filtering or sanitization of the marketplace strings before they are presented to the agent context.
Audit Metadata