audit-component
Skill0 starsChanged 29 days agoMITView on GitHub
Use when asked to review a component, page, or HTML snippet for accessibility, or when an accessibility claim needs evidence attached to it rather than an opinion.
In this repository33 other files it gives its agents.
- AGENTS.mdAGENTS.md
- accessible-authenticationSkill
- accessible-formsSkill
- accessible-nameSkill
- accordion-disclosureSkill
- carouselSkill
- charts-and-datavizSkill
- combobox-autocompleteSkill
- data-tableSkill
- email-html-accessibilitySkill
- fix-color-contrastSkill
- forced-colors-and-dark-modeSkill
- headings-and-landmarksSkill
- keyboard-focusSkill
- live-regionsSkill
- media-captionsSkill
- menu-buttonSkill
- modal-dialogSkill
- motion-and-animationSkill
- page-title-and-languageSkill
- pagination-and-searchSkill
- pointer-and-gesturesSkill
- predictable-contextSkill
- progress-and-loadingSkill
- screen-reader-test-planSkill
- semantic-htmlSkill
- skip-linksSkill
- svg-and-iconsSkill
- tabsSkill
- tooltip-popoverSkill
- triage-automated-resultsSkill
- wcag22-new-criteriaSkill
- zoom-reflow-text-spacingSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Before shipping a component, as a second pair of eyes on markup.
- Distinctive: line 20
When a finding needs to be filed somewhere (a ticket, a PR comment) and a bare "not accessible" is not actionable.
- Distinctive: line 23
Do not use this for a whole site, and do not use it as evidence that something *is* accessible.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Audit a component
Criteria: SC 1.1.1 Non-text Content, SC 1.3.1 Info and Relationships, SC 1.3.2 Meaningful Sequence, SC 1.4.1 Use of Color, SC 1.4.3 Contrast (Minimum), SC 2.1.1 Keyboard, SC 2.4.3 Focus Order, SC 2.4.4 Link Purpose (In Context), SC 2.4.6 Headings and Labels, SC 2.4.7 Focus Visible, SC 4.1.2 Name, Role, Value
Review one component (or one page section) against WCAG 2.2 and produce findings that name the criterion, quote the evidence, and state a fix.
The output is a findings list, not a rewrite. Fixing is a separate step — do not silently refactor the code you were asked to review.
When to use
- "Is this accessible?" / "Review this component" / "Check this page."
- Before shipping a component, as a second pair of eyes on markup.
- When a finding needs to be filed somewhere (a ticket, a PR comment) and a bare "not accessible" is not actionable.
Do not use this for a whole site, and do not use it as evidence that something is accessible. See Failure modes.
Procedure
- Identify what you can actually inspect. From markup alone you can judge structure, names, roles and attributes. You cannot judge focus behaviour, actual computed contrast, or screen reader output. Split the findings into certain and needs runtime check and never merge them.
Walk the component in this order. Structure first, because a wrong structure invalidates everything downstream:
Pass Question Criteria Structure Is this one landmark/heading level, and is its internal order meaningful? SC 1.3.1, 1.3.2 Semantics Does each interactive element expose the right role? SC 4.1.2 Name Does every control, link, image have an accessible name that matches its visible label? SC 1.1.1, 2.4.4, 4.1.2 Appearance Is anything conveyed by colour alone? Are the declared colours sufficient? SC 1.4.1, 1.4.3 Keyboard Is every interactive element reachable, in a sensible order, with a visible indicator? SC 2.1.1, 2.4.3, 2.4.7 Headings and labels Do they describe the content they head? SC 2.4.6 For each finding, emit exactly this shape. A finding without the fix is not a finding, it is a complaint.
[blocking] <one-line what is wrong> criterion: SC 1.1.1 Non-text Content (A) evidence: <line number and the exact snippet> why: <what a user experiences, concretely> fix: <the change to make>Severity is only ever one of:
blocking— fails a Level A or AA criterion. Must be fixed.advisory— passes A/AA but fails AAA, or is a documented best practice.check— cannot be decided from markup. State what to test and how.
- Never invent a criterion. Cite the number and the level. If no criterion
fits, the item is
advisoryor it is not an accessibility issue.
Verify
Re-read each finding and confirm the quoted evidence: line actually appears in
the input. A finding whose evidence was paraphrased is a finding that cannot be
checked.
<!-- Bad: this finding is not checkable -->
evidence: the icon button has no label
<!-- Good: this finding can be verified by the reader -->
evidence: line 3 — <button class="close"><svg …></svg></button>Failure modes
- Reporting a passing audit. Markup inspection cannot establish conformance. Say what you checked and what remains untested.
- Treating "has an aria-label" as passing. An accessible name that does not match the visible label fails SC 2.5.3, and a name of "button" is technically present and useless.
- Duplicating the visible label into a redundant attribute. A
titleon a button that already contains its text adds noise and creates a second source of truth. - Rewriting instead of reporting. The caller asked what is wrong. Larger changes hide which part was the actual defect.
- Flagging
alt=""as a defect. An empty alt on a decorative image is correct. The defect is a missingaltattribute, never an empty one.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

