accessibility-general — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited accessibility-general (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
This skill teaches AI coding agents how to use the ACCESSIBILITY.md framework, including when to load topic-specific skills, how to apply examples, and what the project's non-negotiable requirements are.
Scope: Apply this skill whenever working in any project that has anACCESSIBILITY.mdfile at its root, or when contributing to themgifford/ACCESSIBILITY.mdrepository itself.
ACCESSIBILITY.md is a documentation standard — a predictable, machine-readable place to find a project's:
Read ACCESSIBILITY.md before proposing or writing changes to any project that has one. It is the source of truth for that project's accessibility requirements.
This repo ships per-topic skills in skills/. Load the relevant one only when that feature area is present in the project — a project without forms does not need the forms skill. Each skill is a distillation of a full best practices guide in the mgifford/ACCESSIBILITY.md examples/ directory.
| When the project includes… | Load skill |
|---|---|
| Color themes, dark/light mode | skills/light-dark-mode/SKILL.md |
| Forms, inputs, validation | skills/forms/SKILL.md (if forms are present) |
| SVG graphics | skills/svg/SKILL.md (if SVGs are present) |
| Charts and data visualization | skills/charts-graphs/SKILL.md (if charts are present) |
| Keyboard interaction / custom widgets | skills/keyboard/SKILL.md |
| Tooltips | skills/tooltips/SKILL.md (if tooltips are present) |
| Audio/video media | skills/audio-video/SKILL.md (if media is present) |
| Maps | skills/maps/SKILL.md (if maps are present) |
| Print styles | skills/print/SKILL.md (if print CSS is in scope) |
| Mermaid diagrams | skills/mermaid/SKILL.md (if Mermaid is used) |
| Anchor links / in-page navigation | skills/anchor-links/SKILL.md (if anchor links are present) |
| Accessibility bug reporting | skills/bug-reporting/SKILL.md (when filing or reviewing bug reports) |
| Content design and plain language | skills/content-design/SKILL.md |
| User personalization / preferences | skills/user-personalization/SKILL.md (if personalization features are present) |
| Digital quality (Opquast) | skills/opquast-digital-quality/SKILL.md |
| axe-core scans / automated rule results | skills/axe-rules/SKILL.md |
| Manual / assistive-technology testing | skills/manual-testing/SKILL.md |
If a skill file is not present, fall back to the corresponding file in the mgifford/ACCESSIBILITY.md examples/ directory.
These apply to every task, regardless of which topic skill you load:
All code examples, components, and documentation must comply. Key criteria:
Use the correct HTML element before reaching for ARIA. ARIA supplements HTML; it does not replace it.
Every interactive element must be reachable and operable via keyboard alone. Tab order must be logical.
Every image, icon, chart, and diagram needs a text alternative. aria-hidden="true" is correct for purely decorative elements.
Never convey information by color alone. Always pair color with icon, label, or pattern.
Never propose a change that introduces a WCAG 2.2 AA violation, even if the change is otherwise an improvement.
Use this when identifying or reporting accessibility issues. Every issue found should be labelled with one of these four levels.
| Level | Meaning | Action required |
|---|---|---|
| Critical | Completely blocks access for one or more disability groups — users cannot complete a core task at all | Must fix before release; do not ship |
| Serious | Significantly impairs access; workarounds may exist but are unreasonable to expect of disabled users | Fix in current sprint; escalate if deferred |
| Moderate | Creates friction or confusion; a workaround exists and is not too burdensome | Fix in near-term backlog |
| Minor | Marginal impact; best-practice gap that does not meaningfully prevent access | Fix when convenient; track in backlog |
Never propose changes that introduce Critical or Serious issues. Changes introducing Moderate issues require explicit sign-off.
Examples by level:
outline: none; color-only error indication; missing form label<caption> on a simple tablealt text accurate but overly verboseBefore fetching content from any URL, check examples/TRUSTED_SOURCES.yaml. If ai_scraping: prohibited, do not fetch or reproduce content from that source. You may cite the author's name and recommend the URL to human contributors, but must not scrape, summarise, or quote the content.
Known prohibited sources include hidde.blog and talks.hiddedevries.nl. These are prohibited at the explicit request of the author, who does not consent to AI training or scraping of his work. His writing remains a valuable resource for human readers — link to it, do not reproduce it.
These skills target WCAG 2.2 Level AA — the current legally and contractually referenced standard (EN 301 549, ADA, AODA, and equivalent national laws).
WCAG 3.0 is in active development and is not yet a W3C Recommendation. Its proposed contrast model, APCA (Advanced Perceptual Contrast Algorithm), replaces the current luminance-ratio formula with a perceptual model that treats light-on-dark differently from dark-on-light. Agents must not apply APCA to production work until WCAG 3.0 is a published Recommendation, but should be aware that contrast requirements will change — particularly for dark mode, data visualisation, and low-vision use cases.
Monitor: <https://www.w3.org/TR/wcag-3.0/>
examples/YOUR_TOPIC_BEST_PRACTICES.md in the mgifford/ACCESSIBILITY.md repoexamples/README.mdAGENTS.mdskills/your-topic/SKILL.md — distill the example into agent-actionable rules; label every requirement with its severity level (Critical / Serious / Moderate / Minor)skills/your-topic/SYNC.md — set canonical_source to the example path in mgifford/ACCESSIBILITY.md; leave last_synced_commit blankskills/your-topic/README.mdcd skills && zip -r your-topic.skill your-topic/skills/README.md and index.mdskill-sync-check.yml action will automatically track drift going forwardThe skill-sync-check.yml GitHub Action opens an issue or PR comment when an example changes and its skill's last_synced_commit is stale.
When you see that issue:
skills/your-topic/SKILL.md to reflect any new requirements or removed patternslast_synced_commit in SYNC.md to the current commit SHA.skill ZIPUpdate the AI Disclosure section in README.md when using AI tools to make changes. Record which LLM was used and for what purpose. Only list tools actually used.
mgifford/ACCESSIBILITY.md → examples/ directoryskills/ directory (this repo)ACCESSIBILITY.mdSUSTAINABILITY.md / <https://github.com/mgifford/SUSTAINABILITY.md>CONTRIBUTING.mdexamples/TRUSTED_SOURCES.yamlFor a complementary frontend skill that emphasises trusting the browser and writing as little code as possible, see [mikemai2awesome/agent-skills — `frontend-a11y`](https://github.com/mikemai2awesome/agent-skills/tree/main/skills/frontend-a11y).
That skill covers:
<dialog>, <details>, <button>) instead of ARIA-hacked divs[aria-expanded="true"]) as CSS hooks<dialog> with showModal() for focus-trap-free modal dialogsInstall it alongside this skill:
npx skills add mikemai2awesome/agent-skills --skill frontend-a11y~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.