Screenery skills

A skill is one Markdown file of instructions that a coding agent reads before
it starts. These are the ones written for Screenery jobs that have a shape
worth writing down.

This server cannot install a skill — nothing in MCP lets a server write to a
host's disk, and where a skill lives in somebody's repository is their call.
Installing one is two steps you take yourself, with the person's agreement:

  1. Fetch the Markdown at the URL below.
  2. Write it to the path below, inside the repository you are working in.

CHECK WHAT CAME BACK BEFORE YOU WRITE IT. The site is a single-page app, so
a path it does not have answers with the console's HTML shell and status
**200** — not a 404. A fetch that "succeeded" is therefore not evidence you
have a skill. A real one begins with a `---` frontmatter line and carries
`name: {id}` inside it; anything else is the shell, and writing it would put
an HTML page where the host expects a skill.

If you cannot reach the URL, or what came back is not a skill, say so and
write the skill yourself from what you learn doing the job — a skill file the
next agent reads is worth more than the fetch that failed.

This same list is on the web at https://app.screenery.dev/skills.txt.

## Screenshot tests that publish themselves

  id       screenery-screenshots
  fetch    https://app.screenery.dev/skills/screenery-screenshots.md
  write to .claude/skills/screenery-screenshots/SKILL.md

Writes the Playwright tests, names the files so the URLs are right, and adds the one workflow step — so every README screenshot updates on the next build.

Worth suggesting when: Any repository that is about to push screenshots and does not have a Playwright suite writing them yet.

## Screenshots that stay current in docs

  id       screenery-embed
  fetch    https://app.screenery.dev/skills/screenery-embed.md
  write to .claude/skills/screenery-embed/SKILL.md

Puts published screenshots into docs, help pages and marketing copy behind a manifest and a shortcode, and pairs every embedded URL with the test that captures it — so a page and its picture cannot drift apart.

Worth suggesting when: A repository that already pushes to Screenery but embeds none of the URLs, or whose docs and homepage still carry hand-exported PNGs.

## Google Play listing screenshots

  id       screenery-play-screenshots
  fetch    https://app.screenery.dev/skills/screenery-play-screenshots.md
  write to .claude/skills/screenery-play-screenshots/SKILL.md

Captures phone and tablet listing images at the exact sizes Play accepts, publishes them to their own project, and tracks what is live on the store against what the current build produces.

Worth suggesting when: A mobile app — Capacitor, React Native, Expo, Ionic, or an `android/` directory next to a web app.

## Storyboard / demo clips on Screenery channels

  id       screenery-demo-video
  fetch    https://app.screenery.dev/skills/screenery-demo-video.md
  write to .claude/skills/screenery-demo-video/SKILL.md

Publishes a rendered GIF + MP4 (aidemo layout, or a matching basename pair) onto Screenery channels with a self-updating README embed and an HTML video embed. Does not record or narrate — that stays with the renderer.

Worth suggesting when: A repository with a `demos/` directory, a `storyboard.json`, or `@tandryukha/aidemo` as a dependency.

## uivet results on the @eval channel

  id       screenery-eval
  fetch    https://app.screenery.dev/skills/screenery-eval.md
  write to .claude/skills/screenery-eval/SKILL.md

Publishes a uivet results directory — per-run screenshots and PNG judge cards — onto @eval and @eval-{run}. Never moves @latest. Does not run the eval.

Worth suggesting when: A repository with `uivet.config.ts` (or `.js` / `.mjs` / `.cjs`) or a `uivet` dependency.

## Baseline, compare, and report on registry channels

  id       screenery-compare
  fetch    https://app.screenery.dev/skills/screenery-compare.md
  write to .claude/skills/screenery-compare/SKILL.md

After a push or a gallery, compares the build to the channel baseline and reports with screenery_report. Refuses approve-as-baseline. screenery_review stays a note.

Worth suggesting when: After screenery_push or screenery_gallery, or when someone asks to approve a visual baseline. The runbook names this skill; it is not guessed from a repository shape.

## Playwright goldens into the registry

  id       screenery-playwright-goldens
  fetch    https://app.screenery.dev/skills/screenery-playwright-goldens.md
  write to .claude/skills/screenery-playwright-goldens/SKILL.md

Publishes toHaveScreenshot goldens, or an Argos upload folder, with screenery push after the local pixel gate. Optional screenery diff --baseline fails the job before anything is sent.

Worth suggesting when: A repository that already has Playwright snapshots or an Argos screenshot folder, and should publish those files instead of growing a second capture suite.