The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from teacher transcripts and recording keyframes which could contain adversarial content.
Ingestion points: External data enters the agent context via the read_material and search_material tools as described in SKILL.md.
Boundary markers: There are no explicit instructions to use delimiters or "ignore instructions" warnings when placing extracted transcript evidence into the brief or materialFacts fields of the generate_scene tool.
Capability inventory: The skill is restricted to educational content generation tools (create_stage, generate_scene) and does not possess high-risk capabilities such as arbitrary shell execution, sensitive file access, or outbound network operations.
Sanitization: While the skill requires the agent to build a "compact style profile" and provide "short transcript evidence" (which acts as a summarization filter), there is no explicit sanitization step to check for or remove prompt injection patterns from the source materials.

teacher-style-clone