The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests and processes untrusted code written by users for challenges.
Ingestion points: User challenge solutions inside learn-<topic>/lesson-*/user-*/ are read and analyzed by the verify subagent.
Boundary markers: No specific delimiters or security instructions are specified to prevent the agent from executing instructions embedded within user code files.
Capability inventory: The skill utilizes the Bash tool to execute various shell commands (which diffity, npm install, diffity tree) and verify user-submitted code files.
Sanitization: There is no mention of sanitization, isolation, or sandboxing when executing or reviewing user code.
[COMMAND_EXECUTION]: The instructions command the agent to execute shell utilities to verify tools and manage learning tree instances, specifically calling which diffity and diffity tree --no-open in the background.
[EXTERNAL_DOWNLOADS]: The skill mandates the automated installation of an unversioned package from the public registry via npm install -g diffity if it is not already present on the system.

diffity-learn