zoodata
Pass
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted product reviews from Amazon, creating a surface for indirect prompt injection.
- Ingestion points: Review content is retrieved from external endpoints (e.g.,
/realtime/reviews) and passed into the agent's context through scripts such asscripts/zoodata.py. - Boundary markers: The skill's analysis prompt templates use triple-quote delimiters (
""") to wrap untrusted review text. While this provides a basic separation, it is a weak boundary against adversarial input. - Capability inventory: The agent uses the ingested data to perform sentiment analysis and attribute extraction; it does not directly use this data to trigger sensitive actions like file system modifications.
- Sanitization: The skill does not perform explicit sanitization or filtering of the review text before it is interpolated into the prompts for analysis.
Audit Metadata