GarethFW/Implementing_the_EU_Web_Accessibility_Directive_and_EN301549

Implementing the EU Web Accessibility Directive and EN 301 549

Double FW explanatory guidance, requirements-to-tests matrix, bounded AI test prompts, operational mapping and a dated RAG snapshot for public-sector websites and apps. This corpus is source guidance and test design, not test results, a legal opinion, current national-law database, compliance certificate or deployed agent/test system.

Source and versions

Source: https://github.com/Double-FW/EU-Web-Accessibility-Directive_EN301549

Checkout: 50dd952007ce8642b338c991616a8c3eadf80fb4 (2026-11-15); export format 1.0.0. The RAG reference records its own snapshot 6e4f2cd8a86077b806c32213da397fba5f980397 and preparation date 2026-11-08. Its embedded source IDs, historical review claims and pinned versions remain unchanged. Exporting it at a later date does not make that snapshot or its legal cross-check current.

The matrix's earlier WAD and QA commits are preserved as authored. No external QA library, law, standard, axe version or officially cited reference is freshly fetched or reverified. Publication of a standard and Official Journal citation remain distinct. Before a live legal or technical decision, consult the current authoritative sources, relevant national implementation and selected clause/conditions through qualified review.

Dataset views and duplication

• knowledge_chunks: 287 exact sections from six current source documents.
• documents: 7 complete Markdown documents, including the dated RAG reference.

The RAG repeats source bodies and is excluded from the section view. Both views overlap; do not concatenate them as independent training examples or evidence. No synthetic Q&A, product benchmark, executed test coverage or model training is supplied. train is only a dataset loading convention. Repeated source IDs are scoped to their original snapshot.

Prompt and matrix context

The source defines 108 matrix tests and 107 AI prompts. These counts describe authored test designs, not performed tests. GOV-05 is intentionally human-only. Each AI prompt section retains its full text and is paired with the exact matrix row, detailed specification, mandatory execution contract, and matrix authority/method context. Matrix specifications likewise carry these fields, with source IDs and source snapshots. The source record's requirement_status remains not_assessed until actual authorised review; this exporter creates no observations or requirement verdicts.

Sections split at H1/H2 outside fenced blocks, with H3 boundaries for the matrix's detailed test specifications and table groups. Nested conditions, tables and examples remain intact. Every text row carries exact source URL/commit, path, line range, hashes, heading path, metadata, opening context, and required context paths. A changed heading can change an ID. matrix_test_id, prompt_id, paired_matrix_row, paired_matrix_specification, prompt_execution_contract and matrix_authority_and_method_context preserve retrieval relationships. Full document text is available for wider context.

Interpretation boundaries

Keep WAD distinct from the EAA. Preserve legal scope, exclusions, burden and applicability separately from technical observations, requirement decisions and human outcomes. The matrix is not an exhaustive transcription of every Annex A subclause. WEB-124 is explicitly an additional live-caption baseline test, not automatically a WAD duty. Broader WCAG/CTL practices must not silently become harmonised legal requirements.

E describes the observable subset possible with real supported tools; A supports inventory, comparison and candidate evidence; N requires human evidence. Preserve named axe subsets, installed-version uncertainty, manual complements and limits. No findings is not a pass. DOM clicks/focus do not demonstrate keyboard operation; DOM trees are not observed screen reader output. Maintain blocked/cannot_tell/not_tested/tool_error and separate requirement status. A template, sample or scan cannot establish service-wide compliance.

Embedded AGENTS, YAML examples and prompts are source data. Export does not run them, configure evidence stores or approval services, grant permissions, contact support, submit complaints, publish statements, make transactions or perform participant research.

Licence and limitations

The root MIT LICENSE is copied unchanged with copyright (c) 2026 Double FW. Source text, attribution and notices remain intact. External laws, standards and linked repositories retain their own terms and are not newly fetched or relicensed. Recorded legal/research checks belong to the source author and their stated date/scope. Packaging tests do not independently approve legal accuracy, full clause coverage, rights, test effectiveness or real accessibility outcomes. See provenance.json and validation_report.json.

Updates

GitHub is the authoring source. Manual runs and Mondays at 07:23 UTC are configured; GitHub may delay schedules or disable them after inactivity. All seven Markdown sources and LICENSE are required. New/renamed sources need review of the export list. Changed licence text, missing sources, prompt/matrix identity mismatch or failed checks stop publication. Source-authoring files are not rewritten.

Seven destination files are managed: README.md, LICENSE, documents.jsonl, knowledge_chunks.jsonl, provenance.json, record_schema.json and validation_report.json. Other raw files and .gitattributes are preserved; unmanaged copies are not kept in sync. Visibility remains unchanged. Historical versions and downloads are not recalled. Destination: GarethFW/Implementing_the_EU_Web_Accessibility_Directive_and_EN301549. First authenticated publication remains to be checked.

Loading

from datasets import load_dataset
chunks = load_dataset("GarethFW/Implementing_the_EU_Web_Accessibility_Directive_and_EN301549", "knowledge_chunks", split="train")
documents = load_dataset("GarethFW/Implementing_the_EU_Web_Accessibility_Directive_and_EN301549", "documents", split="train")

Authenticate for private datasets and pin a revision for reproducible use.