Index of every page on this site: llms.txt
All skills
onmax avatar

/before-after-evidence

@aecfada
by Maxonmax/comark-ui43 stars
3
Run it yourself

Shows a visible change as two comparable screenshots, before and after, with marks measured from the source and checked in the rendered page, and a change in motion as a short recording. Use when a pull request, bug fix, or design change alters what people see or how it moves, or when someone asks for before/after screenshots, a recording, or visual evidence.

Use this Skill: https://skilld.dev/gh/onmax/comark-ui/before-after-evidence

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

referencesrecording.md

≈1.6k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Recording motion

A clip is evidence when the reviewer sees the motion at its real speed, starting and ending at rest, with nothing else on screen changing.

1. Script the interaction

Drive the page with a script, such as Playwright, so a retake repeats the same steps at the same pace.

  • Size the viewport to the component, the way you'd clip a still. The viewport is the frame.
  • Rest about 0.7 s before each action and longer than the animation after it, about 1.3 s, so every motion starts and ends on a still frame.
  • Show the motion two or three times, such as three tab switches, and keep the clip to a few seconds.
  • Draw a pointer. A headless browser draws none, and without one the reader sees things change with nothing pressing them. Move the mouse to its target in steps, not in one jump.

2. Capture the frames

Chrome's screencast sends a lossless PNG each time the page paints. Playwright's recordVideo is simpler, but it starts recording during page load and its VP8 encoding blurs small text.

Both capture CSS pixels, whatever deviceScaleFactor says. For sharp text, lay the page out at double size: double the viewport, the root font-size, and any font sizes set in px. Keep CSS zoom and transform off the page. Motion code often measures boxes with getBoundingClientRect, and under either one it draws its moving parts in the wrong place. On a tab transition recorded with zoom: 2, the leaving tokens appeared below the code block.

Save this as record.mjs in a workspace with @playwright/test and its Chromium browser installed, with ffmpeg on PATH. Adjust the viewport and tab names for the page, then run node record.mjs <url> with the page's URL.

import { chromium } from "@playwright/test";
import { execFileSync } from "node:child_process";
import { mkdirSync, writeFileSync } from "node:fs";

const url = process.argv[2];
if (!url) throw new Error("Usage: node record.mjs <url>");

const width = 1456,
  height = 340,
  dir = "clip"; // twice the component's CSS size
mkdirSync(dir, { recursive: true });
const browser = await chromium.launch();
const page = await browser.newPage({ viewport: { width, height } });
await page.goto(url);
await page.addStyleTag({ content: "html{font-size:32px}" }); // plus any px fonts, doubled

// An arrow whose tip follows the mouse, so it leaves the label it points at readable.
const arrow = encodeURIComponent(
  '<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 14 20"><path d="M1 1v16l4-4 3 6 2-1-3-6h6z" fill="#111" stroke="#fff" stroke-width="1.2" stroke-linejoin="round"/></svg>',
);
await page.addStyleTag({
  content: `#pointer{position:fixed;left:0;top:0;z-index:9;width:28px;height:40px;background:url("data:image/svg+xml,${arrow}");pointer-events:none}`,
});
await page.evaluate(() => {
  const tip = document.body.appendChild(
    Object.assign(document.createElement("div"), { id: "pointer" }),
  );
  addEventListener("mousemove", (e) => (tip.style.translate = `${e.clientX}px ${e.clientY}px`));
});

const cdp = await page.context().newCDPSession(page);
const frames = [];
cdp.on("Page.screencastFrame", ({ data, metadata, sessionId }) => {
  frames.push({ data, at: metadata.timestamp });
  cdp.send("Page.screencastFrameAck", { sessionId }).catch(() => {});
});
await cdp.send("Page.startScreencast", { format: "png", maxWidth: width, maxHeight: height });
await page.mouse.move(width - 80, height - 60); // paints the first frame
await page.waitForTimeout(700);
for (const name of ["Second", "Third", "First"]) {
  const box = await page.getByRole("tab", { name }).boundingBox();
  await page.mouse.move(box.x + box.width / 2, box.y + box.height / 2, { steps: 14 });
  await page.waitForTimeout(150); // lets the reader see where it presses
  await page.mouse.down();
  await page.mouse.up();
  await page.waitForTimeout(1300);
}
const end = Date.now() / 1000;
await cdp.send("Page.stopScreencast");
await browser.close();

// A frame arrives only when the page paints, so each one holds until the next.
let list = "";
frames.forEach((frame, i) => {
  writeFileSync(`${dir}/${i}.png`, Buffer.from(frame.data, "base64"));
  list += `file '${i}.png'\nduration ${((frames[i + 1]?.at ?? end) - frame.at).toFixed(4)}\n`;
});
writeFileSync(`${dir}/frames.txt`, `${list}file '${frames.length - 1}.png'\n`);
execFileSync("ffmpeg", [
  "-y",
  "-f",
  "concat",
  "-i",
  `${dir}/frames.txt`,
  "-vf",
  "fps=30,format=yuv420p",
  "-c:v",
  "libx264",
  "-crf",
  "18",
  "-movflags",
  "+faststart",
  "-an",
  `${dir}/clip.mp4`,
]);

The mp4 is H.264, so every browser plays it. A few seconds at this size come to a few hundred KB.

3. Look at the clip

Pull frames from the middle of a motion and from a rest, and look at them:

for t in 0.5 1.1 1.25 1.4; do ffmpeg -y -loglevel error -ss $t -i clip/clip.mp4 -frames:v 1 clip/at-$t.png; done

A frame from the middle of a motion shows what reviewers can't see in a still: parts in between their two places, fading, or drawn where they don't belong.

Done when a frame from the middle of each motion shows the motion your document describes, and the clip starts and ends at rest.

4. Write the document

Keep the shape of a before/after document, with the clip in place of the compare:

Code groups with `magic` now slide the tokens their snippets share into place when you switch tabs. Before, the code swapped all at once.

Open a `::code-group{magic}` and select **tinyjson5**, then **tinyini**.

![Each tab switch slides import and from into place while the package names fade between them](magic-tabs.mp4){width=1456 height=340 autoplay loop muted}
  • The first sentence says what moved before, so the after clip alone carries the comparison.
  • The alt text says what moves, for a reader whose browser can't play the file.
  • autoplay loop muted plays the clip like an animated screenshot. Browsers autoplay only muted video.
  • width and height are the clip's pixels, so the page keeps its room before the file loads. ffprobe -v error -show_entries stream=width,height -of csv=p=0 clip.mp4 prints them.
  • When the change retimes motion that already existed, put the before clip first, recorded with the same script, and title each with a line above it.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

1 warning6d3 checks · Risk SAFE
  • Gen Agent Trust Hub6d

    The skill provides a structured workflow for generating visual evidence (screenshots and recordings) of UI changes. It utilizes standard development tools like Playwright and ffmpeg, and installs utility packages authored by the developer from the npm registry. No malicious patterns or security risks were identified.

  • Socket6d

    No alerts

  • Snyk6d

    Risk: MEDIUM · 2 issues

Signed by skilld at aecfada. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated last week

README badge

README badge for onmax/comark-ui/before-after-evidence