The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external materials such as lesson plans, slide decks, and user-provided concepts, which represents a potential attack surface for indirect prompt injection.
Ingestion points: External data enters the context through the description and 适用范围 (Scope) sections, which allow for the input of user-defined concepts and materials.
Boundary markers: The skill does not explicitly define delimiters to separate untrusted user data from its instructional logic.
Capability inventory: The skill's capabilities are limited to pedagogical content generation using tools like create_stage, generate_scene, and generate_tts. It does not possess capabilities for arbitrary command execution or network exfiltration.
Sanitization: There are no specific instructions for sanitizing or filtering instructions that might be embedded in the uploaded materials.

feynman-learning