Skip to content

role-accessibility-tester.md

Latest commit

 

History

History
275 lines (225 loc) · 15.2 KB

role-accessibility-tester.md

File metadata and controls

275 lines (225 loc) · 15.2 KB
Raw

Role: accessibility tester — recorded NVDA + VoiceOver pass (EN + ES)

Status: alpha / concept stage. habitable is not to be relied on for a real legal matter yet, and one of the few things gating the removal of that caveat is exactly the review described here. This is a paid engagement (or a fair stipend for an individual tester). We do not ask disabled people for free accessibility labor.

We are recruiting one accessibility tester who uses assistive technology daily to do a single bounded, recorded, moderated screen-reader pass of habitable's two user-facing artifacts in English and Spanish, against WCAG 2.2 AA. The findings are published in-repo under our audit-as-artifact model and feed the Accessibility Conformance Report.

This is not an open-ended audit. Automation already does the breadth; we need the one thing automation cannot do — a human who hears the screen reader confirming that the app and the packet are actually comprehensible and operable.

1. Who this is for

The single highest-value tester is a native- or fluent-Spanish screen-reader user, because the load-bearing check is that the reader switches to a Spanish voice when the page language changes — a sighted bilingual tester cannot verify that. We can split the matrix across more than one tester if needed (e.g. one NVDA/Windows user, one VoiceOver/Apple user), but a Spanish-fluent AT user is the one we most want to reach.

You do not need to be a developer or know the codebase. You need to use NVDA and/or VoiceOver as a daily driver and be willing to think aloud while we record the session.

2. Scope — exactly two artifacts, two ATs, two languages

The full protocol you will work from is docs/accessibility/manual-testing.md; findings land in docs/accessibility/ACR.md. The scope below mirrors them.

Two artifacts:

  1. packet.html — the conformant accessible evidence rendering (ADR 0004). Confirm: a single h1; landmarks/headings let you jump between issues and the evidence appendix; the appendix table announces its column headers with each cell; every evidence image has a meaningful text alternative or an adjacent real-text caption; evidence status is announced in words, never by colour alone.
  2. The local web app — uv run habitable app --vault <demo-vault>. Walk the full flow: add an issue → capture a (synthetic) photo → add a timeline entry → resolve timestamps → export → read the result. Confirm: a skip link on first Tab to <main>; labelled controls with required state announced; the resolve button announces its aria-describedby help; results and errors arrive via the polite live region; the language toggle is reachable and operable.

AT matrix (2 artifacts × 2 ATs × 2 languages):

  • NVDA + Firefox on Windows
  • VoiceOver + Safari on macOS, and VoiceOver + Safari on iOS (mobile is in docs/mobile.md scope)
  • English and Spanish, on both artifacts. The Spanish pass must confirm the reader switches to a Spanish voice when the page lang changes — the single most important reason we want a Spanish-fluent AT user.

The EN macOS+iOS VoiceOver run and the ES NVDA+VoiceOver run are the load-bearing ones.

Standard: WCAG 2.2 AA.

What you do NOT need to re-test (CI automation already covers it): axe-core scans in EN + ES, structure/landmarks/alt/labels/no-positive-tabindex, EN/ES string parity, and PWA installability. Your pass focuses on what only a human can judge: real screen-reader comprehensibility, announcement order and quality, reading order, focus management, live-region timing, table semantics in context, image-alt usefulness in context, status-in-words, keyboard operability with no trap, 200% zoom + 320px reflow, prefers-reduced-motion, and the EN↔ES voice switch.

3. Data — synthetic only, never real tenant data

You build everything from generated data; there is no IRB, consent, or privacy friction, and nothing real is ever touched.

uv run habitable demo                        # builds a synthetic demo vault + packet
uv run habitable app --vault <demo-vault>    # the local web app, EN/ES, in a browser/AT

packet.html is produced by habitable demo. Full run instructions are in docs/audits/onboarding.md §2–3. Do not load real tenant data to evaluate habitable.

4. What's provided to you

5. Deliverable

  1. Recordings — screen capture plus audio of the screen-reader speech for each artifact × AT × language combination.
  2. A dated entry for docs/audits/ — your name or handle, the NVDA/VoiceOver/browser versions used, pass/fail per manual-testing.md section, and each issue tagged with its WCAG 2.2 AA success criterion + severity.
  3. Findings that feed ACR.md.

Recording-and-publishing consent is an explicit engagement term. We publish the recording and the report in-repo under our audit-as-artifact model (docs/audits/ + the ACR). Confirm up front that you consent to that, and note any redaction you want (e.g. publish your handle, not your legal name).

Re-test buffer. Per the repo rule, a release must not ship with an open moderate-or-worse manual finding, so the engagement includes a small fix-and-recheck loop: we fix what you find, you confirm the fixes. Please scope a little time for that.

6. Effort estimate (to quote against)

Roughly 1–2 hours of moderated session per (AT × language) combination once setup is done — on the order of a half-day to a full day of tester time for the full matrix, plus the small recheck buffer.

7. Where to find a tester — paid, pro-bono, and community channels

Listed cheapest-realistic to priciest. In every outreach we lead with: it is paid, synthetic-data-only, a public-interest AGPL/Apache FOSS tool, the scope is two artifacts × NVDA+VoiceOver × EN+ES, the deliverable is recordings + a published WCAG 2.2 AA findings list, and we ask explicitly about Spanish-fluent screen-reader testers and whether sessions are recorded by default.

Community — hire an individual (often the most affordable route)

  • A11y / web-a11y Slack (~12k members; free to join, hire individuals). The standard venue to hire a single screen-reader user per session. Join via https://accessibility.github.io/a11yslack/ (the shared invite link rotates; use the welcome page if a direct link 404s). Read the Code of Conduct, post a paid gig in the appropriate jobs/help channel, and ask for Spanish/bilingual screen-reader users.
  • Accessibility Mastodon / Fediverse instances — a11y.social (https://a11y.social/about), dragonscave.space (blind-admin-run, many blind users and AT experts — best place to ask for Spanish-speaking screen-reader users), a11y.info. Post a clear paid gig with image alt text and CamelCase hashtags (#Accessibility #ScreenReader #a11y): scope, NVDA+VoiceOver, EN+ES, recorded deliverable, rate.
  • University Disability Resource Centers / AT labs — especially Hispanic-Serving Institutions and Spanish-language programs for the ES pass. Email AT-lab coordinators describing a paid one-off recorded test on synthetic data and offer a fair per-session stipend. Availability is ad hoc; the synthetic-data-only design removes IRB friction. (e.g. https://accessibility.harvard.edu/Assistive-Technology-Center)
  • National Federation of the Blind (NFB) — Center of Excellence in Nonvisual Accessibility (https://nfb.org/programs-services/center-excellence-nonvisual-access). Formal certification is overkill for an alpha tool, but NFB state/local chapters and member listservs are a strong channel to recruit individual blind testers, including Spanish-speaking members, for a paid per-session pass.
  • AccessAbility Officer — Certified AccessAbility Tester (CAT) program (https://accessabilityofficer.com/blog/cat-program-launch-your-digital-accessibility-career-as-a-blind-professional). Credentialed blind/low-vision testers (~$25/hr). Ask whether you can engage a graduate/apprentice for a single recorded NVDA + VoiceOver pass, and about Spanish capability.

Mission-based vendors — managed, recorded, paid disabled testers (ask for a single-pass minimum and a FOSS/nonprofit rate)

  • WeCo — Digital Accessibility by WeCo (https://theweco.com/accessibility-services/). Disabled certified testers; offers a low-cost/baseline tier and a FREE sample review — a realistic way to scope before committing. Request the free review first (phone 855-849-5050 or the site form), then ask for a baseline disability-focused test on NVDA and/or VoiceOver, recorded, EN+ES. Confirm explicitly that they record sessions, that they cover VoiceOver specifically, and whether any tester works in Spanish.
  • Knowbility AccessWorks (https://knowbility.org/services/accessworks). A managed panel of disabled testers (daily NVDA/JAWS/VoiceOver users), run by a long-standing accessibility nonprofit — engaging them also supports the disability-employment mission. Best fit for the recorded moderated pass. Email aw-services@knowbility.org with the study (artifacts, ATs, EN+ES, recorded sessions + WCAG 2.2 AA findings) and ask for a quote/SOW, the minimum engagement size, whether they record by default, bilingual/native-Spanish tester availability, and any nonprofit/community/FOSS rate.
    • Related: Knowbility AccessU 2026 + the Accessibility Internet Rally (AIR) pro-bono program (https://knowbility.org/programs/john-slatin-accessu-2026/volunteers) — a community to recruit a tester from, and a possible pro-bono structural route. Ask whether a public-interest FOSS legal-aid tool qualifies for AIR or a community-rate engagement.
  • Equalize Digital — Web Accessibility User Testing (https://equalizedigital.com/services/web-accessibility-user-testing/). Real blind/VI testers (partnered with the Texas School for the Blind) using JAWS/NVDA/VoiceOver, paid above minimum wage; values-aligned and approachable for small/independent projects. Submit their contact form for a custom estimate; state the artifacts, ATs (NVDA + VoiceOver, desktop + iOS), EN+ES, that you want recorded sessions, and that it is an unfunded AGPL/Apache FOSS public-interest tool. Ask about a minimum-scope single pass and Spanish capability.

Aspirational — only if a grant or sponsor appears

  • Fable (Make It Fable) (https://makeitfable.com/pricing/). Crowdtesting by disabled daily AT users (NVDA + VoiceOver), with video where you hear what the screen-reader user hears — recordings by default, a direct match to our deliverable. But self-serve Fable Engage starts at ~$3,750 USD/project (verified Jun 2026) — likely out of reach for an unfunded solo project. Treat as the option for if a sponsor/grant appears; if contacting sales, ask whether a smaller single-artifact package or any FOSS/nonprofit discount exists.

Verification note (Jun 2026). Vendor existence, model, and AT coverage are verified. Pricing, minimum scope, whether sessions are recorded by default, and Spanish-fluent tester availability are UNVERIFIED for every agency and must be confirmed by direct contact. Spanish-language coverage is the weakest-confirmed dimension everywhere — the most reliable ES route is hiring an individual native/fluent-Spanish screen-reader user via NFB chapters, dragonscave.space, or an HSI AT lab rather than assuming an agency has one on the bench.

8. Ready-to-send outreach email / post

Subject: Paid recorded screen-reader pass (NVDA + VoiceOver, EN + ES) — open-source tenant-rights tool

Hi —

I maintain habitable, a free, open-source (AGPL-3.0 / Apache-2.0), offline-first tool that helps tenants document housing-habitability problems as tamper-evident evidence. It's an independent, unfunded public-interest project, currently in alpha.

I'm looking to pay a screen-reader user for a small, bounded, recorded moderated accessibility pass against WCAG 2.2 AA. Scope is fixed and small:

  • Two artifacts: an accessible HTML evidence packet (packet.html) and a small local web app.
  • Two assistive technologies: NVDA + Firefox (Windows) and VoiceOver + Safari (macOS and iOS).
  • Two languages: English and Spanish on both artifacts. The most important single check is that the screen reader switches to a Spanish voice when the page language changes — so I'd especially love to reach a native- or fluent-Spanish screen-reader user.

Everything runs on synthetic, generated data — no real personal or tenant data is ever involved, so there's no privacy/consent friction. Estimated effort is roughly 1–2 hours per assistive-technology × language combination (about a half-day to a day total), plus a little time to re-check fixes I make from your findings.

Deliverable: screen + audio recordings of each pass, and a short findings list mapped to WCAG 2.2 AA. I publish accessibility reviews openly in the project's repository as part of how the project keeps itself honest, so consent to publish the recording and report (under whatever name/handle you choose) would be part of the engagement — happy to discuss redaction.

Could you let me know:

  1. Your rate for a pass of this scope (and any minimum)?
  2. Whether you record sessions by default (or whether I should)?
  3. Whether you / your testers can work in Spanish?
  4. Earliest availability?

I'll provide everything needed to run it: the repo, plain-language run instructions, the exact test checklist, and a one-paragraph description of the tool. Thank you — and to be explicit, this is paid work; I'm not asking for free labor.

Best, Chelsea Kelly-Reif — habitable — https://github.com/ChelseaKR/habitable

For Slack/Mastodon, trim to the scope + the four questions + the explicit "this is paid," use the right jobs/gigs channel, and follow each community's Code of Conduct.

9. Related roles

The independent security + cryptographic auditor role and the reviewer onboarding live in docs/audits/onboarding.md and docs/audits/README.md. The housing-lawyer and pilot-partner roles are recruited separately and are out of scope for this brief.