Judgement Tarot — agentic threat model

Not certain from the listing — likely relies on standard commercial LLMs customized via system prompts to adopt different reader personalities. Vulnerable to prompt injection that could bypass the tarot persona, generate offensive content, or leak system instructions.

Not certain from the listing — likely maintains a database of card meanings and user reading histories. Risks include unauthorized access to user chat histories or manipulation of the card database.

Not certain from the listing — likely uses a simple chatbot architecture rather than a complex agentic framework. Risk of tool misuse is low, though interactive card picking must be securely decoupled from LLM generation to prevent manipulation.

Not certain from the listing — deployed as a standard web or mobile application. Vulnerable to typical web application threats such as insecure API endpoints, cross-site scripting (XSS), and denial of service.

Not certain from the listing — likely lacks advanced observability or specialized guardrails, relying instead on the base model's default safety filters to prevent inappropriate psychic advice.

Not certain from the listing — as a freemium entertainment app, it likely lacks rigorous security compliance certifications (e.g., SOC2) or strict data privacy controls beyond standard terms of service.

The agent operates as a standalone vertical application with no described multi-agent coordination or marketplace integrations, resulting in negligible ecosystem risk.