The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to research and process information from external sources to populate a curated list of resources and generate interactive lessons. This process involves a standard indirect prompt injection surface where the agent might inadvertently process instructions hidden within external educational materials.
Ingestion points: The agent ingests data from external URLs and primary sources (books, articles, websites) to create lessons and reference documents stored in MISSION.md and RESOURCES.md.
Boundary markers: The provided markdown templates do not include specific delimiters or 'ignore' instructions to separate untrusted external data from the agent's core teaching logic.
Capability inventory: The agent can write local files (HTML lessons, Markdown records) and execute shell commands to open these files for the user.
Sanitization: The instructions focus on content quality and trust but do not define specific sanitization or filtering protocols for retrieved text before it is interpolated into lessons.
[COMMAND_EXECUTION]: The instructions recommend that the agent use a CLI command to open the generated HTML lesson files for the user (e.g., using open, xdg-open, or start). This is a functional requirement for the user experience within the workspace environment.
[DYNAMIC_EXECUTION]: The skill involves the generation of reusable assets, including interactive quiz widgets and simulators, which are stored as local HTML/JS files in the ./assets/ directory. These scripts are generated based on the learning content to provide a feedback loop for the user.

teach