CSS used to highlight potential performance issues

perf-diagnostics.css

:root {
 --violation-color: red; /* used for clear issues */
 --warning-color: orange; /* used for potential issues we should look into */
}

/* IMAGES */

/*
 * Lazy-Loaded Images Check
 * ====
 * Highlight any lazy loaded images so we can see if any are inside the viewport
 * 
 * Uses an outline so it can pair with Unsized Images and Legacy Format checks
 * Credit: https://twitter.com/csswizardry/status/1346477682544951296
 */
img[loading=lazy] {
    outline: 10px solid var(--warning-color) !important;
}

/*
 * Unsized Images Check
 * ====
 * Highlight images that don't have a height or width attribute set
 * 
 * Uses a border so it can pair with Lazy-Loaded and Legacy Format checks
 */
img:not([height]), img:not([width]) {
    border: 10px solid var(--violation-color) !important;
}

/*
 * Legacy Format Check
 * ====
 * Highlight tiff's and bmp's because we can do better 
 * Also JPG's because maybe we can use something like webp or avif instead
 * 
 * Use opacity so we don't conflict with Lazy-Loaded and Unsized Images checks
 */
img[src*='.jpg'],
img[src*='.tiff'],
img[src*='.bmp']{
    opacity: .5  !important;
}

/* SCRIPTS */

/* Synchronous Scripts Check
 * ====
 * Display any blocking synchronous scripts
 * 
 * Credit: https://twitter.com/csswizardry/status/1336007323337285633
 */
head,
script[src] {
    display: block;
    border: 10px solid var(--violation-color);;
}
    /*
    * Display the URL/filepath of external scripts
    */
    script[src]::before {
        content: attr(src);
        font-size: 1rem;
    }

    /**
    * Hide other head content and non-blocking scripts
    */
    head *,
    script[src][async], script[src][defer], script[src][type=module] {
        display: none;
    }

Took me a while, but here's an interesting solution. Probably longer than I wanted but gives nice, in-place detection for images. These are steps that are needed to make this work:

1. Add the following to the stylesheet (OPTIONAL):

picture:after {
  content: '⚠️';
  margin: -1em;
  font-size: xx-large;
}

2. Add the following JS to the page:

document.addEventListener('DOMContentLoaded', async () => {
  const isFirefox = navigator.userAgent.includes('Firefox');
  const isChrome = navigator.userAgent.includes('Chrome');
  const isSafari = navigator.userAgent.includes('Safari');
  let accept = [];
  let headers = {
    'User-Agent': navigator.userAgent,
  };

  if (isFirefox) {
    const firefoxVersion = navigator.userAgent
      .split('/')
      .pop()
      .split('.')[0];
    if (firefoxVersion >= 65) {
      headers = {
        ...headers,
        Accept: 'image/webp',
      };
      accept.push('image/webp');
    } else {
      accept.push('image/jpeg');
    }
  }

  if (isChrome) {
    headers = {
      ...headers,
      Accept: 'image/webp',
    };
    accept.push('image/webp');
  }

  if (isSafari && !isChrome) {
    accept.push('image/jp2');
  }

  const getImageData = async (image) => {
    const response = await fetch(image.src, {
      method: 'HEAD',
      headers,
    });

    const contentType = (await response.headers
      .get('content-type')
      .includes(';'))
      ? await response.headers.get('content-type').split(';')[0]
      : await response.headers.get('content-type');

    const data = {
      contentType,
    };
    return data;
  };
  const images = document.getElementsByTagName('img');

  const wrap = (elementToWrap) => {
    const wrappingElement = document.createElement('picture');
    elementToWrap.parentNode.appendChild(wrappingElement);
    return wrappingElement.appendChild(elementToWrap);
  };

  for (let image of images) {
    if (!accept.includes((await getImageData(image)).contentType)) {
      wrap(image);
      image.style.opacity = 0.5;
    }
  }
});

The wrap function is optional. Since there's no way to use the :after pseudo-selector for an img element, we need to wrap all imgs to a picture element. This is only a convenient way to differentiate with the "default" opacity behaviour specified in the original gist.

Attached is an example for an image that is served from Cloudinary:

<img
  src="https://res.cloudinary.com/tamas-demo/image/upload/w_500/woman.jpg"
  alt="photo of a woman"
/>

This renders the warning, since the image is viewed in Chrome but it has a JPG extension. Adding f_auto to the URL, will render a WebP:

<img
  src="https://res.cloudinary.com/tamas-demo/image/upload/w_500,f_auto/woman.jpg"
  alt="photo of a woman"
/>

Please remember however that some Image CDNs do analyse the image and it may be more optimal to load a JPEG as opposed to a WebP image in Chrome as well.