amazon-listing-optimization
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill utilizes MCP tools to retrieve data from external platforms including Amazon (product data, reviews, Alexa prompts) and WIPO (trademark/patent records). These connections are documented as necessary for generating listing optimizations and performing IP compliance checks.
- [CREDENTIALS_UNSAFE]: Instructions for setting up the PANGOLINFO_API_KEY are provided, including an example of a JWT-encoded token header. The skill correctly advises users to manage these keys through environment variables or configuration files, adhering to standard development practices rather than hardcoding actual secrets.
- [PROMPT_INJECTION]: The skill identifies an attack surface for indirect prompt injection by ingesting data from external Amazon listings and reviews. Ingestion points: Data is ingested via get_amazon_product and get_amazon_reviews tools. Boundary markers: The skill uses a Data Ingestion Guard to check for data completeness but does not implement specific prompt delimiters for untrusted content. Capability inventory: The agent has access to network tools and text generation capabilities. Sanitization: No explicit sanitization of ingested text strings is described, though the risk is considered inherent to the functional requirement of analyzing listing text.
- [SAFE]: The skill demonstrates safe operational practices, including comprehensive error handling for various failure codes and budget management to prevent excessive resource consumption.
Audit Metadata