WCAG conformance review

When to use it

Use this skill when you need a criterion-by-criterion answer: an audit, input for an ACR (accessibility conformance report) or its VPAT template, or a go-live acceptance check. It covers WCAG 2.1 and 2.2 at levels A and AA.

The assessment gives you the scope, the sampling, a status for each criterion (Pass, Fail, Needs evidence or Not applicable), the evidence and a retest. To fix how a widget behaves, use wp-devkit-accessibility-review.

Where it stops

• This skill decides status against the criteria for a defined scope and keeps the evidence trail. It does not fix components. Each Fail goes to wp-devkit-accessibility-review, with the criterion attached, so that skill can build or repair the control.
• It gives no legal opinion. Name the standard and version your client commits to. Whether laws such as the ADA, the EAA or Section 508 apply is context to confirm with counsel (standards-and-legal-context.md).
• For a question such as "is this dialog accessible" or "why can't I tab here", use the accessibility skill. Use this one when you need a status for each criterion, a sampled audit, a conformance statement or an acceptance gate.

Run a review in Claude

With the plugin loaded, start a read-only review of a bounded target:

/wp-devkit:wcag-review <target-path>

The short form /wp-devkit:wcag runs a quick scan.

Run a review in Codex, ChatGPT or Antigravity

Name the installed skill and keep the request small. Replace the target with your files or journey.

$wp-devkit-wcag-review
Review the target without editing files.
Read the project instructions and the skill's engineering contract first.
Separate confirmed findings from candidates and state the evidence and checks you could not run.

The engineering contract is the set of rules every DevKit skill follows. It defines how findings are classified and how severe each one is.

Reference files

The agent loads only the file that matches your task.

• audit-method.md: used to build the sample and collect evidence.
• wcag-criteria.md: used as the test index of WCAG 2.1 and 2.2 A and AA criteria.
• deliverable-templates.md: used to shape the final report.
• standards-and-legal-context.md: used when the question involves standards versions or legal frameworks.
• wcag-review-workbook.md: used for decision rules, false positives and acceptance cases. The agent reads it first.

How a review runs

1. The agent defines the claim: WCAG version, level, scope, and the accessibility-supported baseline, meaning the browsers and assistive technology you count on. It records exclusions and third-party boundaries.
2. It builds the sample with audit-method.md. The sample has a structured set of templates, components and states, the complete processes a user follows, and a small random set.
3. It picks criteria from wcag-criteria.md for the chosen version. WCAG 2.1 has 50 A and AA criteria including 4.1.1. WCAG 2.2 has 55, and 4.1.1 is removed. The agent opens the normative text and its exceptions before it judges.
4. It gathers evidence for each criterion, using what that criterion needs: source code, the rendered DOM, automated tests, keyboard use, zoom, reflow and text spacing, contrast measurements and assistive technology.
5. It assigns a status with a reason. Pass needs cited evidence. Fail needs reproduction steps. Needs evidence means the criterion was not tested or the result was unclear. Not applicable needs a reason tied to the scope. A clean automated scan or silence never counts as a Pass.
6. For each Fail it records the user impact, severity, owner, a direction for the fix and a retest on the same build. deliverable-templates.md shapes the report.

What you get back

The report opens with the scope and the claim: version, level, sample, test matrix, build and date. One sentence gives the overall result. It says "compliant" only when every in-scope criterion has a Pass with evidence.

A criterion table follows, with the criterion, version and level, status and a reference to the evidence. Each Fail entry lists the criterion, the journey and state, the file and line or URL, steps to reproduce, affected users, impact and severity, confidence, a direction for the fix and the retest. Needs evidence stays separate from Fail, and the report lists technologies, processes and assistive tools that were not tested.

A sampled pass is a pass for that sample only.

Where to go next

Use review to fix to turn findings into a bounded change. Use Doctor and reports for environment and evidence tooling.