The Agent Skills Directory

[SAFE]: The skill is entirely instructional, consisting of design principles, scoring rubrics, and checklists for user interface polish. It does not include any scripts, shell commands, or network-bound operations.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data in the form of UI components and screen frames. Ingestion points: UI rendering and frame-by-frame capture (SKILL.md). Boundary markers: The instructions do not define specific delimiters for external content. Capability inventory: The skill guides the agent in reviewing and modifying code/styles. Sanitization: There are no explicit instructions for sanitizing text or data found within the processed UI. This represents a standard attack surface for agents processing external content, but the risk is assessed as safe given the lack of high-privilege tool access within the skill's scope.

charles-polish