A sub-application is a reusable app or workflow component that can be called inside a larger AI-driven process. In practice, it may handle one job such as search, summarization, ticket creation, or a device command, while the parent assistant decides when to invoke it and what data to pass along.
Sub-applications matter in cyber security because they create another trust boundary. If untrusted prompt content, retrieved text, or user input can influence which sub-application runs, or what arguments it receives, an attacker may redirect the system into unsafe actions. In an agentic platform, that can turn prompt injection into tool abuse, data leakage, or even command execution if a sub-application reaches privileged backends. Defenders reduce this risk with strict allowlists, input validation, permission separation, non-root service accounts, and clear isolation between the model, the workflow engine, and any OS-level action.