The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the primary platform codebase from a repository associated with Tsinghua University (THU-MAIC) and references the OpenClaw project for integration. These are well-known academic and open-source sources.
[INDIRECT_PROMPT_INJECTION]: The system processes user-supplied documents and topics through a LangGraph agent pipeline to generate educational content.
Ingestion points: Document and topic content submitted via the /api/generate endpoint.
Boundary markers: None explicitly defined in the provided instruction set.
Capability inventory: The multi-agent pipeline generates slides, quizzes, and HTML-based interactive simulations.
Sanitization: The platform uses dangerouslySetInnerHTML for custom scene rendering, suggesting it relies on LLM output without extensive HTML sanitization.
[DYNAMIC_EXECUTION]: The CustomSceneComponent uses dangerouslySetInnerHTML to render content generated by the AI pipeline. While this enables the core feature of interactive simulations, it also means that if an LLM is successfully manipulated via indirect prompt injection, it could potentially execute malicious JavaScript in the viewer's browser context.

openmaic-classroom