What problem does it solve? AI-generated UIs frequently ship accessibility defects: low-contrast text, removed focus outlines, mouse-only interactions, and color-only status feedback that exclude users with different vision, motor, and device setups. This Skill gives an AI coding agent concrete rules and a deterministic contrast checker so these defects are caught before delivery. ## Core Features & Use Cases - WCAG Contrast Verification: Ships a Python script (contrast-check.py), the WCAG 2.x luminance formula, and a precomputed reference table so color pairings are computed rather than guessed, with PASS/FAIL verdicts for normal (4.5:1) and large (3:1) text. - Keyboard, Focus, and State Rules: Tell/Why/Fix entries covering focus outlines, tab order, Escape-closable dialogs, non-text contrast (3:1), and perceivable empty/loading/error states. - Mobile Exclusion Checks: Rules for 200% text zoom without clipping and keeping focused inputs visible above the on-screen keyboard. - MCP Server Option: contrast-mcp.py exposes the contrast checker as a Model Context Protocol stdio tool for agents that support MCP. - Use Case: While building a settings page, the agent runs python3 contrast-check.py "#FFFFFF" "#777777", sees normal text FAIL at 4.48:1, and darkens the label before shipping. ## Quick Start Ask the agent to audit your UI's colors and keyboard interactions using the antislop-human checklist and run the bundled contrast checker on every text and background pairing.
