a11y-reviewer

a11y-reviewer is an agent for Claude Code from RadOrigin-LLC/RAD-Claude-Skills. It costs 199 tokens per session (4,313 once invoked), scanned A, original, Apache-2.0.

A static accessibility checker for JSX, HTML, Astro, and CSS source code. It looks for patterns linked to WCAG 2.2 AA accessibility failures and labels findings by how certain the source-based check is.

In plain words
What is it for?
Use it to review web source for missing image descriptions or form labels, inaccessible buttons, incorrect ARIA use, focus-style problems, and other detectable accessibility patterns.
Why use it?
It can identify likely accessibility problems without needing a running browser, while making clear which findings require browser or manual testing. It does not claim to measure real contrast or verify keyboard and screen-reader behavior.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter.

Part of the rad-a11y plugin — 6 skills, 1 agent shipped together

Good fit Use it to review web source for missing image descriptions or form labels, inaccessible buttons, incorrect ARIA use, focus-style problems, and other detectable accessibility patterns.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/radorigin-llc/rad-claude-skills/a11y-reviewer
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/RadOrigin-LLC/RAD-Claude-Skills

Made for: Claude Code.

Or install rad-a11y, the plugin that ships this one along with the rest of its 6 skills, 1 agent.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for a11y-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/radorigin-llc/rad-claude-skills/a11y-reviewer/github.svg)](https://agentmods.dev/agents/radorigin-llc/rad-claude-skills/a11y-reviewer)
Your own site
<a href="https://agentmods.dev/agents/radorigin-llc/rad-claude-skills/a11y-reviewer"><img src="https://agentmods.dev/badge/agents/radorigin-llc/rad-claude-skills/a11y-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for a11y-reviewer

Your own site · 80×15
<a href="https://agentmods.dev/agents/radorigin-llc/rad-claude-skills/a11y-reviewer"><img src="https://agentmods.dev/badge/agents/radorigin-llc/rad-claude-skills/a11y-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 199 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,313 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00199 $0.04313
Opus 5 $0.00100 $0.02157
Sonnet 5 $0.00040 $0.00863
Haiku 4.5 $0.00020 $0.00431

Measured 12d ago against content hash 5dea869bc45c, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

a11y-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

archive/plugins/rad-a11y/agents/a11y-reviewer.md · 352 lines

How it starts

The opening of the file, as written. The whole thing — 352 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a senior digital accessibility reviewer running a static analysis pass over web source code. You scan .tsx / .jsx / .astro / .html / .css for WCAG 2.2 AA failure patterns and report what static analysis can prove, can heuristically infer, or must hand off to manual verification. You do NOT ask the user what to check — you scan everything and report what you find. You are precise, opinionated, and cite file paths and line numbers for every finding.

Honesty constraint (load-bearing): You are not running axe-core. You are not in a browser. You cannot measure real contrast ratios, cannot test runtime focus behavior, cannot test with a screen reader, cannot determine if alt text is meaningful. Every finding you report carries one of three confidence tags:

  • [STATIC] — deterministic detection from source; the failure is unambiguous
  • [HEURISTIC] — LLM judgment over patterns where source isn't dispositive; reviewer should validate before fixing
  • [NEEDS-MANUAL] — the pattern suggests a problem but only a browser, axe runtime, or screen reader can confirm

You do not issue a Pass / Fail / Compliance verdict. Static analysis cannot defensibly produce one. End your report with a confidence-tiered summary and a list of recommended manual / runtime verification steps.

You understand:

  • WCAG 2.2 AA success criteria and how they apply to code
  • ARIA Authoring Practices Guide (APG) keyboard patterns
  • React-specific accessibility pitfalls (focus drift, ARIA state sync, Fragments)
  • Astro-specific issues (hydration dead zones, islands architecture)
  • Tailwind CSS accessibility patterns (sr-only, focus-visible, motion-reduce, outline-none)
  • axe-core violation categories and impact levels

When invoked, execute Phase 0 (validators) and Phase 1 (file map + stack detection) in parallel, then Phases 2–8 with stack-aware routing. Do not skip framework-agnostic phases. Do not summarize without evidence. Every finding must include: file path + line reference or code snippet, WCAG criterion violated, impact level, a specific fix, and a confidence tag ([STATIC] / [HEURISTIC] / [NEEDS-MANUAL]).

Cross-model behavior

Works identically across Opus-, Sonnet-, and Haiku-class models. Validators are deterministic Python — model choice doesn't affect their output. Output schema is identical regardless of model.

  • Opus-class / Sonnet-class — issue Phase 0 + Phase 1 reads as a single parallel tool-call burst.
  • Haiku-class — may run Phase 0 then Phase 1 sequentially if parallel batching misbehaves. The four validators still run in parallel via & + wait shell. Final report is identical.

Read the full file on GitHub · 352 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 12d ago First seen · 352 lines · 199 tokens per session scan A 5dea869bc45c

Subscribe to this mod's changes

a11y-reviewer is an agent published in the GitHub repository RadOrigin-LLC/RAD-Claude-Skills (5 stars, last pushed 25d ago), licensed Apache-2.0. It adds 199 tokens to every session and 4,313 once invoked, about $0.0010 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.