The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions in SKILL.md contain directives that attempt to elevate the priority of its specific rules. It instructs the agent to treat its instructions as the user's own words if they conflict with other system defaults or skills. This is a pattern used to influence the model's instruction-following hierarchy.
[INDIRECT_PROMPT_INJECTION]: The skill transforms untrusted user content into artifacts with embedded scripts, which presents a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context when it reads user-provided research, reports, or content to generate pages as specified in SKILL.md.
Boundary markers: The instructions do not define delimiters or specific warnings to ignore embedded instructions within the processed text.
Capability inventory: The skill possesses the capability to write HTML artifacts containing CSS and JavaScript, which include functionality to interact with localStorage for theme persistence.
Sanitization: There are no instructions to sanitize, filter, or escape the external content before it is interpolated into the generated artifact code.

slide-it