Skip to content
Senthira

Accessibility statement

This page follows the structure of the French déclaration d’accessibilité defined by RGAA 4 — the shape a French accessibility statement is expected to have — and is written in English to match the rest of the site. It says what works, what does not, and what you can do about it.

statement drawn up 6 August 2026 · self-assessment · no third-party audit

Our commitment

Senthira is one person building evaluation tools for AI agents. We want this site to be usable by everyone, including people who navigate by keyboard, use a screen reader, magnify the page, or need motion to stop.

We publish this statement voluntarily. The legal obligation in article 47 of French law n° 2005-102 of 11 February 2005 applies to public-sector bodies and to large companies; Senthira is neither. We follow the RGAA shape anyway, because it is the clearest way to make a claim you can hold us to.

State of conformance

senthira.com is partially conformant with WCAG 2.2 level AA and with RGAA 4.

Partially conformant means parts of this site do not fully meet the standard. We do not claim full conformance. We do not publish a conformance percentage either, because no criterion-by-criterion audit has been carried out — a score we produced about our own work would not tell you anything. What we can give you instead is the specific list of what we know is wrong, in section 05.

Test results

No independent audit has been performed, so there is no RGAA compliance rate to report and no audit report to link to. Everything in this statement comes from a self-assessment: automated checks plus manual review, done by the person who wrote the site.

How this statement was prepared

Drawn up on 6 August 2026, from a self-assessment of the pages listed below. It will be re-dated when the site changes.

Technologies used

  • HTML5
  • CSS (Tailwind CSS)
  • JavaScript / React (Next.js App Router)
  • SVG
  • JSON-LD structured data

Pages are rendered on the server and sent as HTML; there are no third-party embeds, no advertising, no cookie banner, and no personal data collected on this site.

What the assessment covered

  • Keyboard operability: every interactive control is a native link, button, disclosure or form field — nothing is a clickable div — and each one was checked to be focusable, named, and in the page's reading order.
  • Focus visibility: focus styling reviewed on every control, after removing an outline suppression that had been left on one form field.
  • Colour contrast: ratios computed against the background tokens actually used on the page, not against a nominal white or black, and one grey text token was corrected as a result.
  • Structure: landmarks and heading order reviewed page by page, together with what the animated headline and wordmark expose to a screen reader.
  • Motion: behaviour reviewed for the reduced-motion case as well as the default — looping and moving animations stop when the operating system asks for reduced motion. The decorative ring that trails the mouse pointer was reviewed separately: it is not rendered at all on touch devices or under reduced motion, it never takes pointer events, it never replaces or hides the system cursor, and it hides itself as soon as the keyboard is used so that it cannot sit beside a focus indicator.
  • No JavaScript: the pages were checked to stay readable with scripting off, including the demo transcript in the hero.

No automated or manual test was run with a screen reader by someone who uses one daily. That gap is stated again in section 05, because it limits how much this statement is worth.

Pages assessed

Known non-conformities

Listed without spin, each with the criterion it fails so you can check us. This list is not a promise that nothing else is wrong; it is everything we currently know about.

Defects in the site

  • Two short load animations have no control either

    The headline types itself in and the wordmark resolves out of scrambled glyphs when a page loads. Both are under five seconds, both expose stable text to a screen reader, and both are switched off under reduced motion — so neither is a failure of the criterion. We list them because someone who finds them uncomfortable still has no way to stop them.

    WCAG 2.2 — 2.2.2 Pause, Stop, Hide (RGAA 13.8), borderline

  • The in-browser playground does nothing without JavaScript

    With JavaScript disabled, the playground's text field still renders but typing into it produces no verdict, so a control appears to be available when it is not. The rest of the page — including the hero demo transcript — is readable without JavaScript. The real scorer is a command-line tool in the repository and does not depend on this page.

    WCAG 2.2 — 4.1.2 Name, Role, Value / robustness, in spirit

Limits of the assessment itself

  • No independent audit

    Nobody outside the project has audited this site. There is no criterion-by-criterion RGAA evaluation, which is why this statement carries no compliance percentage: a number we produced about our own work would not mean anything.

  • No testing with people who use assistive technology

    The review was done by the person who wrote the site. It has not been tested by a regular screen-reader, switch-access, voice-control or screen-magnifier user, and their report would beat ours.

  • Parts of WCAG 2.2 have not been systematically checked

    The criteria added in WCAG 2.2 — in particular 2.4.11 Focus Not Obscured and 2.5.8 Target Size — have not been walked through breakpoint by breakpoint. The header is sticky, which makes 2.4.11 a plausible risk we have not ruled out. Layout at 400% zoom has not been verified either (1.4.10 Reflow).

No exemption is claimed on grounds of disproportionate burden (« charge disproportionnée »), and no content on this site falls outside the scope of accessibility requirements.

What we plan to fix

  1. 01Walk 2.4.11, 2.5.8 and 400% zoom through every breakpoint, and fix what that turns up.
  2. 02Give the headline and wordmark load animations a way to opt out that does not require changing an operating-system setting.
  3. 03Get an audit from someone who is not us, and a session with people who use assistive technology daily.

No dates promised, because there is one person here and inventing a deadline would be worse than admitting there isn’t one. A reported barrier moves to the front of this list.

Report a barrier

If something on this site stops you doing what you came to do, tell us. Useful things to include: which page, what you were trying to do, what happened instead, and which browser or assistive technology you were using — the pre-filled message asks for exactly those.

Report an accessibility barrier

That address reaches the person who builds the site; there is no support desk behind it. You will get a reply, and it will say plainly whether we can fix the barrier and what we are doing about it. If email is itself the barrier, an issue on the public repository works too.

If we do not fix it — recourse in France

An RGAA statement is expected to tell you where to go when the people who wrote it do not deliver. If you report a barrier here and do not get a satisfactory answer, you can take it to the Défenseur des droits, the French independent rights authority:

One honest caveat: because Senthira is not one of the organisations article 47 binds, this route may not formally apply to us. We list it because RGAA statements do, and because you should not have to depend on our goodwill alone.

Statement drawn up 6 August 2026 · WCAG 2.2 level AA and RGAA 4, partially conformant · back to the home page