The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to invoke specialized interactive quiz tools such as AskUserQuestion, ask_user_question, or quiz. These calls are defined in references/quiz-ui.md to conduct assessments during the teaching process.
[DATA_EXFILTRATION]: The skill manages student state by reading and writing to a local .alvar/ directory as described in references/learner-files.md. These operations are limited to the local filesystem and are used to persist learning history and preferences between sessions.
[PROMPT_INJECTION]: The instructions in SKILL.md require the agent to read user-defined profiles from .alvar/LEARNER.md. This allows user-provided content to influence the agent's instructional style and pace, which is an intended feature of the skill but creates a surface for potential indirect prompt influence.
Ingestion points: .alvar/LEARNER.md, .alvar/maps/*.md, .alvar/sessions/*.md (mentioned in references/learner-files.md).
Boundary markers: The skill uses markdown headers but lacks specific instructions to ignore embedded commands within the learner files.
Capability inventory: File system read/write access and execution of interactive tool calls.
Sanitization: None; the agent consumes the content to adjust its persona and planning.
[EXTERNAL_DOWNLOADS]: The skill's metadata and references/philosophy.md include a link to a YouTube video that serves as the methodological source. This reference is purely informational and targets a well-known, trusted service.

teach