The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it is designed to gather 'knowledge' from external, untrusted resources (articles, books, websites) and 'wisdom' from community forums. Maliciously crafted instructions embedded in these external sources could potentially influence the agent's output or behavior.
Ingestion points: Untrusted data enters the agent context through the RESOURCES.md file (which links to external web content) and through user-provided learning topics via the argument-hint.
Boundary markers: The instructions do not mandate the use of delimiters, XML tags, or specific warnings to ignore instructions embedded within the external materials being processed.
Capability inventory: The skill allows the agent to write files (HTML lessons, Markdown records) and execute CLI commands, which increases the potential impact of a successful injection.
Sanitization: There are no requirements specified for the agent to sanitize, escape, or validate content retrieved from external URLs before processing it.
[COMMAND_EXECUTION]: The instructions explicitly suggest that the agent 'open the lesson file for the user by running a CLI command.' While this is intended to improve user experience by opening generated HTML lessons in a viewer, it involves direct interaction with the system shell.

teach