accessibility
Skill2 starsChanged 15 days agoMITView on GitHub
Design, implement, or audit accessible user interfaces. Use for keyboard and assistive-technology support, focus, names and semantics, contrast, zoom, motion, forms, media alternatives, or WCAG checks across a user flow.
We split this skill into the rules it gives an agent
9 rules. 0 are common, 9 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 99 edited versions we read keep these.
9
Distinctive
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 17
Use the request and available project evidence.
- Distinctive: line 43
Prefer native controls and the project's established accessible components.
- Distinctive: line 47
Use the applicable platform or WAI-ARIA pattern without copying an incomplete example or adding redundant handlers to controls that already …
- Distinctive: line 50
Keep the product's behavior and visual identity unless correcting the barrier requires a change.
- Distinctive: line 53
Do not hide content, remove an essential action, or disable an accessibility check merely to eliminate a finding.
- Distinctive: line 72
Use the relevant screen reader or assistive technology when verifying its experience: navigation, announcements, action, error, and …
- Distinctive: line 75
Check relevant visual conditions, such as text enlargement, reflow, contrast, effective hit areas, supported themes, and reduced motion.
- Distinctive: line 78
Run applicable existing lint, component, or automated accessibility checks.
- Distinctive: line 101
Do not report a whole flow as accessible based only on its first screen, a source patch, a screenshot, or an automated score.
“Common” is measured: kept by at least 6 in 10 of the 99 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Accessibility
Make the intended task perceivable, operable, and understandable through the supported ways of accessing the interface. Check the complete interaction, including entry, state changes, errors, recovery, and the resulting destination. A component with the right attributes can still prevent someone from finishing.
Establish the task and authority
Resolve the platform, affected flow, current revision, project requirements, design-system components, and supported browser or OS and assistive technologies. Use the request and available project evidence; ask only when an unresolved choice would materially change the behavior or acceptance criteria.
An audit produces findings and coverage. Implement corrections when the active task authorizes them. A screenshot supports visual observations and design advice; it does not establish keyboard behavior, accessible names, or spoken output. An unrelated backend change does not need a UI audit.
For web work, use the project's accessibility target; if absent, state WCAG 2.2 Level AA as the engineering baseline. This is a scoped working target, not a claim of legal compliance or whole-product conformance. Read references/web.md before selecting web checks. For native apps or terminal interfaces, read references/platforms.md and consult the relevant platform documentation. Web attributes, CSS pixels, and platform units are not interchangeable. Confirm the exact criterion and its exceptions before treating a preference, pattern recommendation, or enhanced target as a requirement.
Define the interaction before changing it
Trace the affected task through its relevant states, including loading, empty, invalid input, denied access, interruption, and success where they exist. Identify what users must perceive, what actions they can take, where focus should go, and how they discover the outcome and recover. Include essential embedded or third-party steps in that coverage even when their code is outside our control.
Prefer native controls and the project's established accessible components. Inspect their actual composition and runtime behavior: a library's reputation or an ARIA attribute does not prove this use works. For a custom interaction, resolve its semantics, keyboard model, focus lifecycle, and announcements together. Use the applicable platform or WAI-ARIA pattern without copying an incomplete example or adding redundant handlers to controls that already handle the key.
Keep the product's behavior and visual identity unless correcting the barrier requires a change. Make the smallest coherent repair at the responsible component or shared primitive. Check affected callers when changing a shared control. Do not hide content, remove an essential action, or disable an accessibility check merely to eliminate a finding.
Collect evidence that matches the claim
Choose checks from the affected interaction and required coverage, not a fixed
number of screens or a tool's default scan. Use the host's available UI tools and
the project's existing harness. When applicable, invoke verify with this flow
and its accessibility criteria; the matching verify-<app> owns startup, health,
fixtures, driving, and cleanup. A missing specialist does not prevent source
inspection, but missing execution remains a limitation.
Combine the evidence the claim needs:
- Inspect the composed source and accessibility tree for names, roles, states, relationships, order, and hidden content. Check the meaning, not just presence.
- Drive the real keyboard or platform input path through the task and its transitions. Observe focus entry, movement, visibility, dismissal, and return. A composite widget may use one Tab stop and arrow keys internally.
- Use the relevant screen reader or assistive technology when verifying its experience: navigation, announcements, action, error, and recovery. A tree, synthetic event, or automated scan cannot stand in for that interaction.
- Check relevant visual conditions, such as text enlargement, reflow, contrast, effective hit areas, supported themes, and reduced motion. Record the actual condition; narrowing a viewport is not evidence that browser zoom was used.
- Run applicable existing lint, component, or automated accessibility checks. Inspect their scope, exclusions, inconclusive results, and false positives. An empty violation list proves only what that run could check.
For an observed defect, preserve enough of the failing action and result to compare after the repair. Add a focused regression where the repository's tests can protect the behavior, then rerun the affected path and relevant neighbors on the final revision. Do not replace an unavailable manual check with a passing automation score. State the missing environment or action and the next check.
Report barriers and coverage
For each supported finding, give the location or state, affected user task and input method, reproduction or source evidence, expected and observed behavior, and a concrete correction. Cite the applicable criterion or platform guidance when making a standards claim. Distinguish a confirmed failure, a usability recommendation, and a question that needs execution. Prioritize by blocked tasks, impact, reach, and available alternatives rather than a universal severity label.
State what passed within the tested scope, what failed, and what remains untested, blocked, or inconclusive. Include the revision, relevant environment and settings, checks actually run, and retained evidence. Required missing checks prevent a complete result; third-party ownership does not turn a blocked task into a pass. Do not report a whole flow as accessible based only on its first screen, a source patch, a screenshot, or an automated score.
Compose only where needed
review-code-changeskeeps the review scope and final findings when requesting this specialist. Return evidence in its format without starting another audit.verification-authoringmaintains a reusable local verification recipe when requested; this skill supplies accessibility criteria for that recipe.error-handlingowns failure and recovery semantics; this skill checks whether users can perceive the error and reach the recovery action.design-code-structureowns module and interface boundaries when a repair needs structural design. Ordinary accessibility work does not require it.
Other versions109 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 109
- microsoft/vscode★ 194ksame name
- elastic/kibana★ 21ksame name
- mozilla-mobile/firefox-ios★ 13ksame name
- midudev/preguntas-entrevista-react★ 7.8ksame name
- gitpod-io/openvscode-server★ 6.2ksame name
- posit-dev/positron★ 4.3ksame name
- legions-developer/evilcharts★ 3.2ksame name
- midudev/la-velada-web-oficial★ 1.9ksame name
- UI5/webcomponents★ 1.8ksame name
- theopenco/llmgateway★ 1.7ksame name
99 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
Used alongsideProjects with it also use tool-rename-deprecation, author-contributions and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
gitpod-ioSkilltool-rename-deprecation
Ensure renamed built-in tool references preserve backward compatibility. Use when renaming a toolReferenceName, tool set referenceName, or any tool identifier. Run on ANY change to tool registration code. Covers legacyToolReferenceFullNames for tools and legacyFullNames for tool sets.
Also in 5 projects that use this
MicrosoftSkillauthor-contributions
Identify all files a specific author contributed to on a branch vs its upstream, tracing code through renames. Use when asked who edited what, what code an author contributed, or to audit authorship before a merge. This skill should be run as a subagent — it performs many git operations and returns a concise table.
Also in 8 projects that use this
MicrosoftSkillauto-perf-optimize
Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.
Also in 7 projects that use this
- +9 more from microsoft
In this repository48 other files it gives its agents.
- agent-instructionsSkill
- analyze-change-effectsSkill
- article-processingSkill
- ci-cd-automationSkill
- comment-styleSkill
- commitSkill
- communicate-clearlySkill
- compare-solutionsSkill
- create-project-instructionsSkill
- daily-meeting-updateSkill
- debugSkill
- deprecate-and-removeSkill
- design-code-structureSkill
- domain-searchSkill
- error-handlingSkill
- explain-codeSkill
- explain-decisionsSkill
- github-actionsSkill
- handoffSkill
- humanizeSkill
- issue-workflowSkill
- observabilitySkill
- people-memorySkill
- performance-optimizationSkill
- planningSkill
- pr-followupSkill
- prototypeSkill
- pr-review-followupSkill
- pr-reviewSkill
- prSkill
- report-work-activitySkill
- retroSkill
- review-code-changesSkill
- simplify-codeSkill
- skill-maintenanceSkill
- stacked-prSkill
- task-breakdownSkill
- tddSkill
- teachSkill
- typescript-best-practicesSkill
- verification-authoringSkill
- verifySkill
- wayfinderSkill
- workflow-to-skillSkill
- work-modeSkill
- worktrunkSkill
- write-documentationSkill
- youtube-processingSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.

