The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to search the internet for authoritative learning materials and download them into the local materials/ directory. While this facilitates the skill's primary function, it involves fetching content from non-whitelisted remote sources.
[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/generate_board.py) to process knowledge base data and generate an HTML visualization dashboard. This involves running a script provided with the skill which performs file I/O operations.
[PROMPT_INJECTION]: The skill ingests untrusted data from the internet (search results and downloaded materials) and processes user-provided files. This creates an attack surface for indirect prompt injection, as the agent is instructed to use these materials to guide the learning process without explicit sanitization or instructions to ignore embedded commands. Evidence of attack surface: Ingestion points in materials/ folder; capability inventory includes memocli file access, Python execution, and internet access; no boundary markers or sanitization logic specified.

learning-assistant