The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided design briefs to guide the generation of UI artifacts, which introduces a surface for indirect prompt injection.
Ingestion points: User-supplied briefs processed according to the task modeling steps in SKILL.md.
Boundary markers: No explicit delimiters or instructions are used to separate user data from the core design instructions.
Capability inventory: Writing HTML/SVG files and executing the scripts/design-check.mjs tool which interacts with a browser.
Sanitization: No specific sanitization or filtering of the brief content is described.
[DYNAMIC_EXECUTION]: The scripts/design-check.mjs script performs dynamic execution by using the playwright library to run a headless browser and execute JavaScript via page.evaluate(). This is used to audit generated artifacts for design floors such as text contrast and touch target sizing.
[COMMAND_EXECUTION]: The agent is instructed to run a local Node.js script (scripts/design-check.mjs) to verify its output, which is a routine practice for development and design workflows.
[EXTERNAL_DOWNLOADS]: The verification script has a functional dependency on the playwright and playwright-core libraries. These are standard, well-recognized tools for browser automation.

design