Description-Behavior Mismatch
- Category
- MCP Tool Poisoning
- Confidence
- 87% confidence
- Finding
Profile validation and Schema.org conversion support the declared entity-data purpose. However, the dedicated AI-observation model and summarize command add a distinct capability for measuring entity visibility in captured AI answers, which the description does not disclose. This is a limited scope mismatch, not evidence of unauthorized network access or data exfiltration.
- Content
