Skydis — agentic threat model

Not certain from the listing — The specific foundation models used by Skydis are not disclosed. The primary threats at this layer include adversarial prompt injection to bypass chatbot guardrails and model reprogramming to output malicious content.

The platform allows chatbots to be trained on custom business data. This introduces risks of data poisoning (if malicious data is ingested) and data exfiltration, where attackers use prompt injection to extract sensitive business documents from the RAG knowledge base.

Orchestration relies on a drag-and-drop builder and Zapier integrations. Insecure tool integration is a major threat here; a compromised or manipulated chatbot could trigger unauthorized Zapier webhooks, leading to downstream data leaks or unauthorized actions in connected business tools.

Not certain from the listing — The hosting infrastructure, sandboxing of execution environments, and network security controls are not detailed. Standard SaaS threats like container escape or widget-based Cross-Site Scripting (XSS) on client websites apply.

Not certain from the listing — There is no mention of built-in evaluation frameworks, real-time guardrails, or security observability tools to detect prompt injection or anomalous chatbot behavior.

Not certain from the listing — Compliance certifications (such as SOC 2 or GDPR alignment) and fine-grained access controls for managing the chatbot's training data and integrations are not specified.

Not certain from the listing — While Zapier allows multi-app workflows, there is no native multi-agent ecosystem or marketplace described, limiting direct agent-to-agent trust abuse threats within the platform itself.