ecommerce-walmart-product-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external marketplace product data from Walmart, which could potentially contain adversarial content designed to influence the agent.
- Ingestion points: Data is ingested from API responses via the Nexscope proxy in
scripts/sorftime_walmart_product_analysis.py. - Boundary markers:
SKILL.mdinstructs the agent to summarize only returned facts and preserve missing values as unknown to maintain objectivity. - Capability inventory: The skill script has network access to the Nexscope API and performs local file writes to manage session data and a search cache.
- Sanitization: The script parses responses as structured JSON, but no explicit sanitization or filtering of the text content is performed before the data is summarized by the agent.
Audit Metadata