Fabrice AI — agentic threat model

Not certain from the listing — Fabrice AI is model-agnostic and does not specify a default foundation model, meaning model-level threats like adversarial examples or data poisoning depend entirely on the user's choice of LLM.

Not certain from the listing — Being stateless by default, the framework does not mandate specific vector databases or RAG pipelines, leaving data operations and knowledge-base poisoning risks to the developer's implementation.

As a functional, composable TypeScript framework, L3 is highly relevant. Risks include insecure tool integration, prompt injection bypassing functional constraints, and framework-level vulnerabilities if dynamic code execution or untrusted inputs are passed to composed functions.

Not certain from the listing — The framework is infrastructure-agnostic, meaning deployment security, container sandboxing, and secrets management are completely decoupled from the framework and must be managed externally.

Not certain from the listing — The minimal, dependency-light description does not mention built-in evaluation, logging, or guardrail mechanisms, leaving observability gaps unless integrated manually.

Not certain from the listing — As an open-source, minimal framework, there are no mentioned built-in identity, authorization, or compliance controls (such as SOC2 or ISO alignment).

With a strong focus on 'team collaboration' and composability, the agent ecosystem layer is critical. Threats include cascading failures across composed agents, trust abuse between collaborative agents, and insecure delegation of tasks within the agent network.