amazon-review-intelligence-extractor
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill communicates with the vendor's API at api.zoodata.ai to retrieve product and review insights. The script includes a security check in the
api_callfunction ofscripts/zoodata.pythat verifies the destination host before transmitting the API key, ensuring the Bearer token is only sent to trusted ZooData hosts or localhost. This prevents credential leakage to unauthorized servers. - [PROMPT_INJECTION]: The skill is designed to process untrusted data from Amazon reviews, which presents an indirect prompt injection surface. The skill manages this risk through structured prompting in its fallback workflow.
- Ingestion points: Raw Amazon reviews are fetched via the
reviews-rawsubcommand inscripts/zoodata.pyand passed to the agent for analysis. - Boundary markers: The tagging and reduction prompts defined in
scripts/zoodata.pyuse triple-quote delimiters (""") to encapsulate review content and prevent instructions from bleeding into the agent's context. - Capability inventory: The skill possesses network access to
api.zoodata.aiand the ability to execute the local Python scriptzoodata.py. - Sanitization: The prompts explicitly instruct the agent to output only valid JSON and to translate non-English content before extraction, focusing on specific data dimensions to minimize the impact of any adversarial content within the reviews.
Audit Metadata