---
name: a11y-quick-scan
description: Keyboard-path and contrast sweep of named pages with fixes ranked by how many users they block. Use before shipping a page or flow, or when an accessibility complaint arrives - a fast, concrete pass that produces small fixes landable in the same PR, not a WCAG dissertation.
---

# A11y Quick Scan

## When to use

- Before shipping a new page, form, or flow
- After any report that someone "can't use" part of the product

Do NOT present this as a full WCAG audit. It is a quick scan of the
named pages; say exactly what was and was not covered.

## Pass 1: Keyboard only

Put the mouse away. For each named page:

1. Tab through start to finish. Every interactive element must be
   reachable in a sensible order, with a visible focus indicator.
2. Operate everything with keyboard alone: menus, modals, dropdowns,
   date pickers, drag lists.
3. Traps and escapes: a modal must trap focus while open and release
   it on close; Escape must close what looks closable.
4. Note every stop where focus vanishes, order jumps, or a control
   cannot be operated.

## Pass 2: Contrast and meaning

1. Body text and interactive labels: check contrast against their
   actual background, including hover, disabled, and placeholder
   states. Flag pairs below 4.5:1 (3:1 for large text).
2. Meaning without color: errors marked only in red, required fields
   marked only by color, charts readable in grayscale.
3. Names and labels: icon-only buttons need an accessible name;
   inputs need a label that is not just placeholder text.
4. Non-text contrast: focus indicators and the visual boundaries of
   interactive components need 3:1 against adjacent colors (WCAG
   1.4.11). A focus ring you can barely see fails Pass 1's "visible"
   requirement - judge it against this threshold.

## Rank by users blocked

Rank every finding: a broken keyboard path on the checkout button
outranks a low-contrast footer link. For each: what is wrong, who it
blocks, and the smallest fix that unblocks them - each small enough
to land in the same PR.

## Output format

Findings ranked, one per line: element and page, the failure, the
fix. Then a coverage note: pages scanned, passes completed, known
gaps (screen-reader pass, zoom to 200%, motion) not covered here.

## Hard boundaries

- Never claim compliance; claim what was tested and what passed.
- Never fix contrast by changing brand colors silently - propose the
  change, name the rule it satisfies.
- If automated tooling was used, verify its flags by hand before
  reporting them; tool noise is not a finding.
