Back
T
TEMPLATESCAN TEMPLATESCAN

TemplateScan

Multi‑engine accessibility audits with statistically defensible whole‑site coverage

TemplateScan is a multi‑engine accessibility scanning stack built to answer a hard question for large websites: how many pages are actually affected by each WCAG issue, not just which pages you sampled.

lera-cloud/lera-scan

97

Hours

3

Weeks

283

Commits

Solo Developer

Role

Private
Work in Progress
About

TemplateScan is a multi‑engine accessibility scanning stack built to answer a hard question for large websites: how many pages are actually affected by each WCAG issue, not just which pages you sampled. It runs multiple audit engines against authenticated and unauthenticated experiences, reconciles their findings into a single axe‑compatible report, then uses template clustering and sampling to estimate whole‑site impact. This lets product teams, compliance leaders, and vendors see a clear picture of platform versus third‑party issues and make decisions based on honest, quantified accessibility risk.

Under the hood, TemplateScan is organized as three npm packages in TypeScript: a browser‑free core for normalization and WCAG crosswalk logic, a Playwright‑based engine layer, and a discovery module for crawling sitemaps and sitewide links. The core defines a canonical axe‑shaped violation schema, reconciles multiple WCAG versions and best‑practice rules into consistent tags, and performs cross‑engine deduplication using WCAG success‑criterion signatures combined with a browser‑agnostic DOM identity system that stamps stable selectors across frames and shadow roots. The engine package coordinates axe, QualWeb, Alfa, IBM, HTML_CodeSniffer, and Testaro, applies authenticated session capture and replay with secret redaction, and enforces a never‑throw contract so long‑running scans over thousands of URLs can complete with per‑engine error records instead of crashing.

To turn sampled pages into trustworthy “affects N pages” statements, TemplateScan applies MinHash and locality‑sensitive hashing to cluster URL templates, uses accessibility skeleton fingerprints to avoid over‑ or under‑merging layouts, then runs stratified reconciliation with Wilson confidence intervals and section‑aware extrapolation. This yields governance‑ready metrics that distinguish native platform patterns from vendor widgets, support benchmarking different settle strategies, and make multi‑engine accessibility reporting practical at enterprise scale.

This system powers multi-engine accessibility scanning, turning heterogeneous tool outputs into statistically grounded, axe-compatible sitewide WCAG compliance insights for product, legal, and vendor governance teams.

Key Features

Engineered a canonical axe-compatible violation schema with strict XPath exclusion and provenance validation so multi-engine findings merge safely into existing scoring and UI pipelines.

Built a cross-engine deduplication layer keyed on WCAG success-criterion tags and stable DOM node identifiers to eliminate double counting while preserving per-engine attribution and impact severity.

Implemented a browser-agnostic DOM identity system that stamps elements with stable selectors across frames and shadow roots, enabling reliable multi-engine merging and repeatable attribution for large-scale audits.

Designed a WCAG standards scoping and crosswalk framework that reconciles multi-version criteria, obsolete success criteria, and best-practice rules into consistent axe-style tags with explicit gap logging.

Developed a template clustering and representative sampling engine using URL templates, MinHash locality-sensitive hashing, and accessibility skeleton fingerprints to scan fewer pages while preserving statistically defensible coverage.

Created a stratified reconciliation model using Wilson confidence intervals and section-aware extrapolation to convert sampled page results into honest whole-site ‘affects N pages’ estimates for governance reporting.

Implemented vendor fingerprint attribution that tags third-party widgets via class and ID patterns while keeping WCAG scoring pure, providing nuanced platform-versus-vendor visibility without distorting compliance metrics.

Integrated authenticated session capture and replay with heuristic login detection and secret redaction so accessibility scans accurately reflect real logged-in experiences without leaking credentials in errors.

Built with
TypeScriptJavaScriptaxe-corePuppeteerPlaywrightQualWebSiteimprove Alfaaccessibility-checkercommandercsv-writerpuppeteer-extratestaroxml2jscheerioHTML_CodeSniffer
Impact

This system powers multi-engine accessibility scanning, turning heterogeneous tool outputs into statistically grounded, axe-compatible sitewide WCAG compliance insights for product, legal, and vendor governance teams.