The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill contains examples for processing real-time events received via WebSockets from Zoom. Any skill that ingests external data has a potential surface for indirect prompt injection if that data is processed by an LLM without strict boundaries or sanitization. However, the provided examples focus on logging and internal application logic.
Ingestion points: The WebSocket message event handler in SKILL.md and references/connection.md receives data directly from the Zoom service.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the code snippets.
Capability inventory: The provided scripts demonstrate logging to the console and basic status updates; no file system writes, subprocess executions, or external network transmissions are performed within the event handlers.
Sanitization: The implementation uses JSON.parse to handle the incoming data but does not show explicit sanitization of the payload content before use.
[External Reference]: The documentation includes a link to an external GitHub repository (just-zoomit/zoom-websockets) as a sample implementation. Users are encouraged to review the source code of any third-party repository before deploying it in their own environment.

setup-zoom-websockets