Cross Page Analyzer

cross-page-analyzer

Internal sub-agent. This agent is not user-invokable. It is orchestrated automatically by web-accessibility-wizard (and the accessibility-lead team) when auditing multiple web pages. You do not need to invoke it directly.

What It Does

cross-page-analyzer receives aggregated scan findings from multiple web page audits and:

1. Classifies issues — determines whether each finding is systemic (affects all pages), template-level (affects pages sharing a component), or page-specific (isolated to one page)
2. Scores each page — computes a weighted 0–100 accessibility risk score with an A–F grade
3. Generates a cross-page comparison scorecard — a table showing all pages side-by-side with scores and issue counts
4. Tracks remediation progress — when a baseline is provided, classifies each finding as Fixed, New, Persistent, or Regressed

When It Runs

The web-accessibility-wizard calls this agent during multi-page audits:

• audit-web-multi-page prompt — after scanning 2+ pages
• compare-web-audits prompt — when comparing two audit reports across time
• Full site audits triggered by the accessibility-lead in a comprehensive review

Scoring Formula

The agent computes a risk score for each page:

Page Score = 100 − (sum of weighted findings)
Minimum score: 0

Severity	Confidence	Deduction
Critical	High (confirmed by both axe-core and code review)	−15 points
Critical	High (single source)	−10 points
Critical	Medium	−7 points
Serious	High	−7 points
Serious	Medium	−5 points
Moderate	High	−3 points
Moderate	Medium	−2 points
Minor	Any	−1 point

The double-source multiplier for Critical issues (+5 deduction) reflects that axe-core findings confirmed by manual code review have very high confidence and almost always represent real barriers for users of assistive technology.

Grade Table

Score	Grade	Meaning
90–100	A	Excellent — meets WCAG AA
75–89	B	Good — mostly meets WCAG AA
50–74	C	Needs Work — partial compliance
25–49	D	Poor — significant barriers
0–24	F	Failing — unusable with assistive technology

Issue Classification

The agent examines which issues appear across multiple pages and assigns each a type:

Type	Definition	Fix Strategy
Systemic	The same issue appears on every audited page	Fix the shared layout template, navigation component, or global CSS — highest ROI
Template-level	The issue appears on pages that share a specific component	Fix that shared component
Page-specific	The issue is unique to one page	Fix individually

Why this matters: A systemic issue fixed in <Header /> or <Footer /> disappears from every page at once. Fixing the highest-impact systemic issues first produces the greatest accessibility improvement per hour of engineering work.

Comparison Scorecard

For a 5-page audit, the agent produces a scorecard table like:

Page	Score	Grade	Critical	Serious	Moderate	Minor
Home	72	C	2	3	1	4
Product detail	88	B	0	2	1	2
Checkout step 1	61	C	3	2	0	3
Checkout step 2	58	C	3	3	1	2
Account settings	91	A	0	1	0	1

This scorecard appears in the audit report and in commit comments when the wizard posts results to a pull request.

Remediation Tracking

When baseline report data is provided, every finding is classified:

Classification	Meaning
Fixed	Present in baseline, absent in current scan
New	Absent in baseline, present in current scan
Persistent	Present in both scans
Regressed	A previously Fixed issue has returned

The compare-web-audits prompt surfaces this classification to users as a progress overview: "12 Fixed, 3 New, 8 Persistent, 1 Regressed."

Connections

Component	Role
web-accessibility-wizard	Orchestrating wizard; calls this agent after page scans and incorporates output into the audit report
audit-web-multi-page prompt	User-facing prompt that triggers this agent for multi-page comparison
compare-web-audits prompt	User-facing prompt that triggers remediation tracking analysis
web-severity-scoring skill	Scoring formulas and grade table this agent implements