The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external subject material, which is a common attack vector for indirect prompt injection. However, it explicitly mitigates this in references/subject-research.md by instructing the agent to treat any embedded instructions within sources as content to be analyzed rather than commands to be followed.- [DYNAMIC_EXECUTION]: In references/subject-research.md, the skill suggests verifying educational code examples by testing them in an 'authorized harmless environment.' This is a standard and controlled practice for ensuring the accuracy of teaching material.- [DATA_EXFILTRATION]: The skill includes strong privacy guardrails, particularly in references/subject-research.md, where it explicitly forbids sending private learner work to external search services without direct permission.- [EXTERNAL_DOWNLOADS]: The skill references several research papers and university-hosted abstracts in references/pedagogy-evidence.md. All referenced domains (e.g., wustl.edu, cmu.edu, ucla.edu, ies.ed.gov) belong to well-known and trusted academic or governmental organizations.

teach-me