interface-design
Pass
Audited by Gen Agent Trust Hub on Nov 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions and documentation (in
references/checks.md) describe the execution of Node.js scripts to automate UI audits. These scripts (scripts/ui-check.mjsandscripts/ui-modes.mjs) use the Playwright library to interact with a web browser, take screenshots, and write result files to the local filesystem. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external content from web page UIs provided via a URL. This creates an attack surface where instructions embedded in the target website's text or metadata could attempt to influence the agent's auditing conclusions.
- Ingestion points: UI text content, CSS properties, and accessibility trees of the target URL processed by
ui-check.mjsandui-modes.mjs. - Boundary markers: None explicitly defined in the instructions, although the agent is constrained by a structured heuristic checklist in
references/standards.md. - Capability inventory: The skill can execute Node.js scripts, write files (screenshots, JSON results) to the local filesystem, and perform network navigation via a browser context (implemented in the
scripts/directory). - Sanitization: No content sanitization is performed on the UI text before it is counted or reviewed by the agent.
- [DYNAMIC_EXECUTION]: The
scripts/ui-modes.mjsfile utilizes dynamic module loading via theimport()function to load a user-provided tasks file specified by the--tasksargument. This is a standard design for extensible testing tools.
Audit Metadata