F/MS Startup Game — agentic threat model
The F/MS Startup Game is an educational platform with low-to-moderate agentic risk, primarily posing threats related to intellectual property leakage of user startup ideas and potential prompt injection bypassing game boundaries.
OWASP AIVSS score rationale
| Autonomy of Action | 0.30 | |
| Goal-Driven Planning | 0.40 | |
| Self-Modification | 0.10 | |
| Dynamic Tool Use | 0.30 | |
| Persistent Memory | 0.50 | |
| Contextual Awareness | 0.40 | |
| Dynamic Identity | 0.20 | |
| Multi-Agent Interactions | 0.50 | |
| Non-Determinism | 0.60 | |
| Opacity & Reflexivity | 0.40 |
Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.
MAESTRO 7-layer threat model
Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.
Not certain from the listing — likely relies on commercial LLMs to power the AI co-founders. Primary threats include prompt injection to bypass game constraints or extract underlying system prompts, and model utility denial.
Not certain from the listing — likely maintains a database of user-submitted startup ideas, market research, and progress. Vulnerable to unauthorized data access or leakage of user intellectual property.
Not certain from the listing — orchestrates game state and progress tracking. Vulnerabilities could allow users to manipulate game state variables or bypass validation steps to artificially progress.
Not certain from the listing — hosted as a closed-source web application. Standard web application vulnerabilities (e.g., broken authentication, cross-site scripting) represent the primary infrastructure threats.
Not certain from the listing — no public details on guardrails. Gaps in observability could allow the AI co-founders to provide highly inaccurate, hallucinated, or inappropriate business and legal advice without detection.
Not certain from the listing — as a freemium educational tool, it likely lacks rigorous enterprise compliance frameworks (e.g., SOC2) or robust data deletion guarantees for user-submitted business ideas.
Features a 'selection of AI co-founders' to guide the user. This multi-persona ecosystem is vulnerable to conflicting agent instructions, persona-breakout attacks, and cascading logic failures if co-founders interact or evaluate each other's feedback.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).