The Agent Skills Directory

[SAFE]: The skill operates entirely within a local course directory structure, using a strictly defined set of allowed and forbidden file paths to prevent unauthorized data access.
[DATA_EXPOSURE]: The instructions explicitly forbid reading sensitive files like roadmap.md or .audit/ and restrict writing to a single specific file (.records/office-hours.md), ensuring proper data isolation.
[COMMAND_EXECUTION]: No command-line tools or shell executions are invoked; the skill relies purely on text analysis and response generation.
[EXTERNAL_DOWNLOADS]: There are no network operations, URL fetches, or external package dependencies present in the skill.
[PROMPT_INJECTION]: The skill includes behavioral boundaries (the "wall") that prevent it from providing implementation help or homework fixes, even if requested by the user, which acts as a safeguard against behavioral manipulation.

office-hours