A11y Review Coach

Automated WCAG 2.2 AA review, fix, and impact analysis — powered by IBM Bob.

Impact Summary

Lighthouse Accessibility Scores — Before & After

Page Score Change

Findings Explorer

How It Works

  1. Define the skill

    A structured a11y-review skill captures the full WCAG 2.2 AA review workflow — discover, classify, fix, report — in a single reusable instruction file Bob loads on demand.

    Bob feature: Custom Skills (.bob/skills/)

  2. Parallel subagent review

    Three subagents run simultaneously — one per page — each applying the skill and returning a structured findings object. Total review time: ~10 minutes.

    Bob feature: Parallel Subagents (spawn_subagent)

  3. Apply fixes

    Bob makes targeted, surgical edits to HTML, CSS, and JS for every High and Medium finding, logging a before/after code snippet for each change in FIX_LOG.md.

    Bob feature: Agent Mode with apply_diff / search_and_replace

  4. Verify & measure

    Lighthouse audits are run manually before and after; detection rate is computed against the ground-truth issue list and written into metrics.js for this dashboard.

    Bob feature: Structured data contracts (window.A11Y_METRICS)

  5. Report

    Findings and metrics are serialised into findings.js and metrics.js, loaded here without any build step or CDN — GitHub Pages serves everything as plain static files.

    Bob feature: Plan Mode + write_file for structured artefacts