The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to process and display untrusted data from external files and conversation history, which creates a surface for indirect prompt injection.
Ingestion points: User-named files (e.g., walk me through subtasks.json), conversation history, and inline lists.
Boundary markers: The skill does not define specific delimiters or warnings to ignore instructions embedded within the processed items.
Capability inventory: The skill allows file modification via the edit command.
Sanitization: No explicit sanitization or filtering of the ingested content is mentioned.
[COMMAND_EXECUTION]: The edit command allows the agent to modify files or rewrite content based on the item currently under review. This is a core functional capability intended for interactive workflows like code refactoring or task updates.

walkthrough