This gives Claude a weighted scoring formula to turn accessibility findings into 0-100 scores with A-F grades, where critical issues from both axe-core and agent review hit hardest at 15 points each. You'd use it when you need consistent math across audit reports, not just a list of problems. It handles confidence levels (high/medium/low), tracks remediation between audits (fixed/new/persistent/regressed), and classifies whether issues are systemic across all pages or page-specific, which helps prioritize fixes. The grading rubric is opinionated but practical: 90+ is WCAG AA compliant, anything below 50 means significant barriers. Good for teams running repeated audits who need apples-to-apples comparison and want to show progress over time.
npx skills add https://github.com/community-access/accessibility-agents --skill web-severity-scoring