The Agent Skills Directory

[PROMPT_INJECTION]: The skill incorporates untrusted user input via the $ARGUMENTS variable into the agent's context. This creates a surface for indirect prompt injection where a user-provided task description could contain instructions designed to bypass the intended formatting logic or influence subsequent agent behavior.
Ingestion points: The $ARGUMENTS placeholder in SKILL.md receives the user's task description.
Boundary markers: The input is wrapped in <task_description> XML-style tags, providing some structural separation.
Capability inventory: The skill itself does not define any subprocess calls, file-write operations, or network access.
Sanitization: No explicit sanitization, validation, or escaping of the user-provided input is performed before it is interpolated into the prompt.

generate-task