Suspicion Agent — agentic threat model
Suspicion Agent is a specialized gaming AI utilizing GPT-4 for strategic, theory-of-mind reasoning in imperfect information games. Its primary risks are confined to game-state manipulation and local execution vulnerabilities typical of open-source research code, presenting low real-world threat.
OWASP AIVSS score rationale
| Autonomy of Action | 0.70 | |
| Goal-Driven Planning | 0.80 | |
| Self-Modification | 0.10 | |
| Dynamic Tool Use | 0.20 | |
| Persistent Memory | 0.30 | |
| Contextual Awareness | 0.80 | |
| Dynamic Identity | 0.10 | |
| Multi-Agent Interactions | 0.70 | |
| Non-Determinism | 0.60 | |
| Opacity & Reflexivity | 0.70 |
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.
Uses GPT-4 as its foundation model. Vulnerable to prompt injection that could manipulate its strategic decision-making or force it to reveal hidden game state information.
Not certain from the listing — likely manages game state history and opponent modeling in-context, but any external storage of game logs or player profiles could be vulnerable to state poisoning.
Not certain from the listing — the orchestration framework managing the Theory of Mind reasoning loop and game-action execution is unspecified, but flaws here could allow out-of-order execution or state desynchronization.
Not certain from the listing — as an open-source project, deployment is likely local or self-hosted, exposing the host to standard dependency vulnerabilities or local code execution risks if the game environment is compromised.
Not certain from the listing — observability is likely limited to standard game console logging, leaving potential blind spots regarding adversarial prompt injections from other players.
Being an open-source gaming project, there are no formal enterprise security controls, compliance alignments, or strict access management policies in place.
Operates in a multi-agent or agent-to-human ecosystem inherent to imperfect information games. Vulnerable to collusion, social engineering by other agents, or exploitation of its Theory of Mind logic by adversarial players.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).
These scores are auto-generated from public information (the agent's own listing, docs, and repository) using the canonical OWASP AIVSS formula and the MAESTRO framework — an estimate for guidance, not a penetration test, audit, or certification. See the scoring methodology. Are you the vendor? Factual corrections are free.