The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill incorporates external user input into the agent's primary instruction set through the use of the $ARGUMENTS variable.
Ingestion points: External content is loaded via the $ARGUMENTS placeholder at the end of SKILL.md.
Boundary markers: The variable is enclosed within <instructions> tags, providing a structural delimiter.
Capability inventory: The skill is configured to use the Write tool (which can modify the local filesystem), AskUserQuestion, and the Agent tool for orchestration.
Sanitization: There is no evidence of input validation, escaping, or sanitization performed on the content of $ARGUMENTS before interpolation.

interview