The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from a variety of external and local sources, including web-based research, reference materials, and files from a user's knowledge vault (e.g., Obsidian). This ingestion workflow creates a surface for indirect prompt injection attacks if the source materials contain adversarial instructions intended to influence the agent's behavior during lesson generation or record keeping.
Ingestion points: RESOURCES.md, ./reference/*.html, ./learning-records/*.md, and documents/PDFs from the configured Obsidian knowledge vault.
Boundary markers: Absent; there are no specific instructions to use delimiters or ignore embedded instructions when reading these files.
Capability inventory: The skill performs file system writes (HTML and Markdown), executes CLI commands to open files, and invokes research tools.
Sanitization: No explicit sanitization or validation logic is defined for data retrieved from external sources before it is interpolated into lessons or records.
[COMMAND_EXECUTION]: The skill instructions direct the agent to run CLI commands to open lesson files for the user (e.g., using open or similar tools). While this is a functional feature for the tutor's primary purpose, it involves executing shell commands based on generated file paths.
[EXTERNAL_DOWNLOADS]: The skill utilizes research tools to fetch knowledge and wisdom from external sources. The instructions emphasize the use of high-trust resources, such as recognized experts and peer-reviewed work, and specifically mentions well-known services like Reddit for community interaction. These references are documented neutrally as they are integral to the skill's educational purpose.

teach