The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes npx playwright to capture screenshots of live websites for analysis. This command execution is scoped to the specific task of UI auditing and is a standard procedure for web-based design reviews. The execution is limited to specific viewport and screenshot parameters.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, including website HTML, CSS, and Figma metadata. This ingestion represents a potential indirect prompt injection surface.
Ingestion points: Website HTML/CSS, Figma metadata (text, components, variables), and pen.dev app states are ingested via Playwright, Figma MCP, and pencil MCP respectively.
Boundary markers: The skill employs a highly structured analysis framework (nine dimensions of a 'Design Read' and specific RYUX Knowledge categories) which serves as a logical delimiter, guiding the agent to evaluate specific UI/UX attributes rather than following arbitrary text content.
Capability inventory: The skill utilizes command execution for screenshots, external tool invocation for design metadata, and heuristic evaluation tools.
Sanitization: No explicit sanitization of text nodes or HTML content is defined, though the instructions focus the agent's attention on visual and structural analysis rather than content execution.

ryux-critique