devs Tools / Quality Gate 1.0.0 / WCAG 2.2

Accessibility Studio: Audit, remediate, and document WCAG 2.2 checks

Audit HTML and CSS in an isolated local workspace, separate automated, assisted, and manual WCAG 2.2 A/AA checks, apply reversible safe candidates, re-audit, and export twelve evidence-aware formats.

Accessibility Studio is a free, local-first audit workspace. It separates machine-detectable findings, assisted review, and human-only checks; preserves evidence and confidence; offers reversible safety-A candidates; and exports twelve report and handoff formats.

  • Source, findings, and ZIP generation stay in this browser
  • Opaque-origin preview with network, forms, and objects blocked
  • No conformance, certification, or legal guarantee
108Rule contracts 51Runtime checks 30Manual checks 12Outputs 日本語

First visit?Choose a fixture, run an audit, inspect evidence, record manual outcomes, then export. A zero-finding result is not a conformance claim.

View terminology

Opening the isolated local audit workspace…

Evidence workflow

From source to evidence, remediation, re-audit, and handoff

01

Choose a safe input

Start with one of ten fixtures or paste HTML/CSS without scripts or external resources.

02

Set the environment

Compare viewport, text zoom, color scheme, and reduced-motion conditions.

03

Review evidence

Read stable IDs, targets, evidence, confidence, severity, and human-review boundaries.

04

Fix and re-audit

Apply only reversible safety-A candidates, inspect the diff, undo if needed, then audit again.

05

Record manual checks

Record pass, fail, N/A, or not tested with device and assistive-technology evidence.

06

Export with limits

Export one of twelve formats while keeping untested items and conversion limits visible.

Decision boundary

Automation can find signals; people determine meaning and usability

Automated checks alone do not establish WCAG conformance. WCAG 2.2 A/AA is the audit baseline, not a certification result. AAA criteria may appear as advisory context. Final evaluation requires representative pages, real devices, keyboard use, screen readers, zoom and reflow, and documented human judgment.

AutomatedDeterministic signals such as a missing attribute.
AssistedEvidence is collected, but a person confirms context.
ManualMeaning, operation, quality, and assistive-technology behavior.

Ten audit fixtures

Compare broken, accessible, and mixed examples

Open a practical guide

Evidence, confidence, and fix safety

Read why a finding exists before changing code

FieldMeaningUse
EvidenceThe DOM, computed styles, selectors, and measurements used for a decision.Confirm that the target and evidence are reproducible.
ConfidenceHow reliable the decision is, separate from impact.Treat low-confidence results as hypotheses for human review.
Safety AA narrow candidate that can be applied to the working copy and undone.Review the diff and scope, then re-audit.
Safety B / CA fix that requires content or structural judgment.Do not send the candidate directly to production.
Safety NAn item that is never fixed automatically.Record the reviewer, environment, outcome, and evidence.

Twelve output adapters

Handoff the same evidence without hiding loss

Compare all outputs
AdapterBest forPreserved information
JSON audit reportMachine processing and re-auditsFindings, evidence, environment, and manual outcomes
Markdown audit reportReviews and issuesReadable issues and remediation guidance
HTML executive reportBrowser reviewA self-contained summary and evidence
CSV FindingsSpreadsheets and trackingOne finding per row
SARIF 2.1.0CI and code scanningRule IDs and locations in a standard format
CSS remediation diffStyle fixesCandidates with safety level and scope
WordPress Rescue handoffWordPressTargets and fix candidates for the existing tool
Section Composer Quality ContractPage compositionSection IDs and quality conditions
Design Token Contrast ContractColor systemsColor pairs, ratios, and text conditions
GSAP Motion Alternative ContractMotion designPin, scrub, and final-state requirements
WCAG-EM Evidence PackEvaluation recordsEvidence that is not a conformance statement
QA RunbookPre-release checksDevices, assistive technology, steps, and results

The WCAG-EM evidence pack supports evaluation work; it is not a conformance statement.

Public knowledge base

Inspect every public contract behind the audit

108 ChecksStable Rule IDs, target criteria, detection conditions, evidence, limitations, fixes, and safety. 10 PatternsARIA, keyboard, focus lifecycle, motion, mobile, and implementation QA. 59 WCAG criteriaMappings between WCAG 2.2 criteria and devs Rules, with AAA advice kept separate. 12 OutputsPurpose, media type, mappings, preserved data, installation, and removal. Practical guidesKeyboard, zoom and reflow, WordPress, screen-reader, and motion procedures.

Glossary

Terms used in the audit workspace

WCAG
International guidelines for making web content usable by more people. This tool uses WCAG 2.2 A and AA as its baseline.
Success Criterion / SC
A testable requirement used to evaluate WCAG conformance, such as 1.4.3 Contrast (Minimum).
Finding
A candidate issue found by a specific rule from the target code and supporting evidence.
Evidence
Selectors, attributes, computed styles, ratios, and viewport data used to reproduce a finding.
Confidence
How reliable a detection is. It is separate from severity.
Audit Project
The canonical data that combines input, environment, findings, manual checks, fixes, re-audits, and exports.
Remediation Patch
A candidate fix and its before-and-after diff. Only safety-A patches can be applied to the working copy.
Re-audit
Running the audit again after a fix to check the same conditions and detect regressions.
Reflow
Content rearranging at narrow widths or high zoom without horizontal scrolling or information loss.
Accessible Name
The name assistive technology announces for a control such as a button or input.
SARIF
A standard JSON format for sending analysis findings to CI and code-scanning systems.
WCAG-EM
A method for evaluating representative pages and workflows across a website. An evidence pack alone does not establish conformance.

Security and privacy

What stays local and what is intentionally blocked

  • HTML, CSS, audit results, and manual notes are not sent through URLs, analytics, or external APIs.
  • The outer and inner iframes do not receive same-origin privileges; CSP blocks network access, forms, objects, base URLs, and external resources.
  • Scripts, event handlers, srcdoc, form actions, and external resources are removed. Messages are checked against source, token, audit ID, and schema version.
  • HTML is limited to 250 KB and CSS to 150 KB. Files are created only after an explicit action.

Known limits

Checks that still require people and real environments

  • Alt-text quality, heading and link meaning, reading order, and understandable error messages.
  • Completing tasks with a keyboard, focus trapping and restoration, and content revealed on hover or focus.
  • Text at 200%, reflow at 320 CSS pixels, text spacing, and multiple browsers and devices.
  • Assistive technology such as NVDA and VoiceOver, plus WordPress theme and plugin combinations.

Frequently asked questions

Questions before relying on the result

Does zero findings mean WCAG 2.2 AA conformance?

No. It means the automated and assisted checks run in this environment detected no issue. Evaluation still requires the 30 manual checks, representative pages, real devices, and assistive technology.

Is pasted HTML or CSS sent to the server?

The public workspace processes it in the browser and blocks external communication with CSP. Only the optional WordPress admin reference plugin API sends data explicitly to the same logged-in site.

Are multiple h1 elements automatically a violation?

No. A person must review the page topic, heading hierarchy, and how assistive technology communicates the structure.

Should buttons and links be 24px or 44px?

WCAG 2.2 Level AA Target Size (Minimum) generally requires 24 by 24 CSS pixels or the defined spacing exception. devs also shows 44 by 44 CSS pixels as a usability recommendation, not as the AA minimum.

Can a safety-A fix be published automatically?

No. Candidates apply only to the isolated working copy, with a diff, Undo, and re-audit. Production changes remain a separate step after backup and review.