The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses open or xdg-open to display generated HTML lessons to the user. This is a standard method for presenting rich content and does not involve high-privilege operations.
[DATA_EXPOSURE]: The skill reads and writes to several files in the current workspace (MISSION.md, GLOSSARY.md, RESOURCES.md, etc.) to maintain state across sessions. This is the intended behavior for a stateful teaching tool.
[PROMPT_INJECTION]: There is a potential surface for indirect prompt injection as the agent is instructed to read instructions and state from workspace files (MISSION.md, NOTES.md, learning-records/) to determine its behavior. If an attacker provides these files, they could influence the agent's teaching logic. However, the skill explicitly mandates grounding lessons in a curated RESOURCES.md file to mitigate reliance on untrusted or parametric knowledge.

teach-me