Audit or remediate web accessibility barriers, optimizing accessible task completion against a WCAG 2.2 AA floor.
The English summary is being prepared. Showing the original description.
Run score Not tested
See how to installReview the instructions the agent will receive before you install.
Optimize accessible task completion within an explicit scope, with WCAG 2.2 Level AA as the minimum target where conformance is in scope. Automated scores are supporting evidence, never the objective. Limit every claim to the surfaces, states, environments, and methods actually tested.
Route the run before anything else:
Scope is one of: a changed feature or component, selected routes and journeys, or a representative product-wide sample. A sampled audit claims the sample, never whole-product WCAG conformance.
Resolve mode, target surfaces, representative user tasks, authentication and data constraints, supported browser and assistive-technology combinations, available tooling, and where transient evidence lives. Prefer a narrower explicit scope over an implied whole-product claim.
Done when mode, scope, support baseline, constraints, and evidence location are all explicit.
Key every row route-or-journey × state × viewport × input/AT mode. Cover representative routes, shared layouts and primitives, essential journeys, responsive variants, and materially different UI states — authenticated, empty, populated, loading, success, validation-error, permission-error, dialog, menu, disclosure, toast, drag — where they exist in scope.
Done when every selected row has a stable identifier.
Detect and reuse the repository's package manager, start command, browser tests, accessibility libraries, and reporting conventions. Run automated checks against rendered UI in every matrix row, then load references/manual-audit.md and complete every applicable category.
For axe WCAG 2.2 A/AA coverage, tags are incremental — select all five:
wcag2a,wcag2aa,wcag21a,wcag21aa,wcag22aa
Retain violations, incomplete results, exclusions, tool versions, browser, state, and viewport per row. Keep raw tool output transient unless the user asks to commit a report.
Prefer tools in this order:
Prefer project-local or ephemeral execution over global installs.
Done when every matrix row has automated and applicable manual evidence, or an explicit untestable reason.
Prioritize in this order: essential-task blockers, reach across shared surfaces, confirmed WCAG A/AA impact, user impact, confidence, then fix leverage versus regression risk. Task evidence overrides raw axe impact.
Done when every finding has an affected matrix scope, user consequence, evidence, standard or best-practice basis, confidence, proposed remedy, and verification method. Audit mode proceeds directly to step 7.
Fix root causes in small verifiable batches. Prefer native HTML, visible labels, and established project primitives; treat every ARIA role as a keyboard and focus contract. Keep every rule enabled, and record any exclusion with its impact on the claim. Add or update regression tests in the existing test stack.
Done when every in-scope confirmed A/AA failure and critical task barrier is fixed, or explicitly deferred with a reason and evidence.
Rerun the identical matrix. Reload references/manual-audit.md before declaring completion. Compare axe evidence with scripts/compare-evidence.mjs and inspect every new or incomplete result. Exercise essential journeys with keyboard and each declared browser/assistive-technology combination — this is where every ARIA contract from step 5 is honored or exposed.
Done when no unexplained matrix mismatch, new finding, or unresolved incomplete result remains; applicable manual checks and project tests pass; and remaining limitations are recorded.
Report mode and scope, tested environment and versions, findings or changes, per-row evidence, unresolved risks, exclusions, and claim limits. Axe zero and Lighthouse 100 are instrument readings, not conformance.
Done when, in audit mode, every finding is reported and prioritized with its evidence and claim limits and product source is unchanged; in remediate mode, every matrix row shows before/after evidence or a documented untestable reason, with deferrals, untested areas, and enhancements beyond AA listed.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Dissect a codebase and run a 10-question mock technical interview on it, rating each answer 1-10 and closing with a PASS or FAIL verdict.
The English summary is being prepared. Showing the original description.
gostaria que você agora usasse a força total do mega brain com.... o cop thief. que está acionado no sistema. ele precisa investigar mais… termos inteligentes e o que está acontecendo agora, na atualidade do mercado para que esse vídeo tenha tantos acessos e visualizações quanto o primeiro vídeo. o antigo. ele precisa ser uma versão reformulada e eu gostaria que você ajustasse somente aquilo que achar necessário para… se conectar ao contexto inicial que eu colei no começo da conversa. precisamos de um vídeo inteligente. didático, profundo na voz do tiago finch, porém voltado para empresários.
The English summary is being prepared. Showing the original description.
Audit a codebase for dead and otherwise safe-to-remove code, and report each finding for approval. Use when asked to find dead or unused code, deduplicate repeated constants, replace magic literals that duplicate a named constant, or remove arbitrary caps, thresholds, or timeouts.
The English summary is being prepared. Showing the original description.
Flag comments that merely restate the code instead of explaining why, and remove only those the user approves. Use when asked to remove dumb, redundant, or obvious comments, clean up comments, or invoke `/remove-dumb-comments [<number>|all]`.
The English summary is being prepared. Showing the original description.