Woz — agentic threat model

Not certain from the listing — The underlying foundation models are not specified, leaving the platform vulnerable to standard LLM risks such as prompt injection, adversarial reprogramming, or model-inherent biases that could affect code generation.

Not certain from the listing — There is no detail on how user business ideas, proprietary code, or operational data are stored, partitioned, or protected against data exfiltration and knowledge-base poisoning.

Not certain from the listing — While the agent performs software development and bug fixing, the orchestration framework is undisclosed, raising concerns about insecure tool integration and potential execution of malicious code during the development lifecycle.

Not certain from the listing — The hosting, sandboxing, and secrets management mechanisms for executing generated code and managing third-party integrations are not described, presenting risks of container escape or privilege escalation.

Not certain from the listing — No evaluation, monitoring, or logging guardrails are mentioned, which could lead to blind spots in detecting anomalous agent behavior or malicious code injections.

Not certain from the listing — The platform is closed-source and paid, but does not cite specific compliance certifications (e.g., SOC2, ISO) or identity and access management controls governing the generated business infrastructure.

Not certain from the listing — The platform integrates with third-party business tools and services, but the security of these external connections and the potential for cascading failures across the integrated ecosystem remain unverified.