The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides an interface for processing untrusted external data, which is an inherent part of realtime communication applications.
Ingestion points: Hooks such as useChat, useTranscriptions, and useTextStream defined in references/data-hooks.md ingest data from remote participants.
Boundary markers: The skill does not provide specific boundary enforcement, as it is a library of hooks; sanitization is left to the implementing application.
Capability inventory: The hooks allow for UI state changes, audio/video track management, and data channel communication.
Sanitization: React's default protection against XSS during rendering is present, but content processing for LLMs would require additional application-level guardrails.

react-hooks