The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes methods for ingesting and displaying real-time data from remote users, such as names and avatar URLs, which constitutes an untrusted data ingestion surface.
Ingestion points: Data is accessed via the useCursorUsers() hook (rules/shared/data/data-cursor-hooks.md) and the componentConfig.cursorUser template object (rules/shared/wireframe-variables/wireframe-variables-cursors.md).
Boundary markers: The instructions do not explicitly require the use of delimiters or "ignore" instructions when the agent renders user-supplied strings.
Capability inventory: The skill's documented capabilities are limited to UI rendering and configuration; no dangerous capabilities like arbitrary shell execution or file writes are associated with the processing of this data.
Sanitization: The guide focuses on implementation patterns and does not explicitly detail sanitization steps for user-supplied data.
[DYNAMIC_EXECUTION]: The skill documents the use of framework-specific template directives for runtime evaluation of component configuration.
The rules/shared/wireframe-variables/wireframe-variables-cursors.md file describes the use of directives such as velt-if, velt-data, and velt-class for dynamic UI state management. This is categorized as simple script generation from known templates, which is a standard feature of the Velt UI framework.

velt-cursors-best-practices