The Agent Skills Directory

[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks (Category 8) as it processes untrusted content from live URLs, component code, and design files.
Ingestion points: Untrusted data enters the agent context via live URLs, screenshot evaluation, and analysis of component code or DESIGN.md files (SKILL.md).
Boundary markers: The instructions do not establish clear delimiters or "ignore embedded instructions" warnings for the UI content being audited.
Capability inventory: The skill utilizes browser automation tools for DOM inspection (evaluate_script), visual capture (take_screenshot), and system-level port detection (lsof).
Sanitization: Textual content extracted from the audited interfaces is used for evaluation without explicit sanitization to mitigate instructions hidden in UI text.
[COMMAND_EXECUTION]: The skill suggests the use of the lsof command to detect active local development server ports. This is a standard developer diagnostic tool used to facilitate navigation to the correct local environment.
[EXTERNAL_DOWNLOADS]: The skill uses npx to run the @google/design.md utility for linting design tokens. This is a reference to an official tool provided by a trusted organization (Google) and is documented as a core part of the design review process.

ui-design-critique