Skip to content
Back to search

preview-lab

MarketplaceWorks with Claude CodeGitHub

Preview Lab: fast UI/UX screen review for coding agents.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 20 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/dbalas/preview-lab with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/dbalas/preview-lab'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/dbalas--preview-lab/badge.svg)](/listings/dbalas--preview-lab)

HTML

<a href="/listings/dbalas--preview-lab"><img src="/api/listings/dbalas--preview-lab/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

npx skills add dbalas/preview-lab

Then ask your agent:

let me review the new checkout screens

That's it. The agent captures the screens, opens the lab in your browser, and waits for you to press **Send**.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 14 lines
  1. {
  2. "name": "preview-lab",
  3. "description": "Preview Lab: fast UI/UX screen review for coding agents.",
  4. "owner": {
  5. "name": "dbalas"
  6. },
  7. "plugins": [
  8. {
  9. "name": "preview-lab",
  10. "source": "./",
  11. "description": "Fast UI/UX screen review for coding agents: capture, review in a branded lab, feedback straight back to the session."
  12. }
  13. ]
  14. }
README.md · README · 145 lines
  1. <p align="center">
  2. <img src="docs/logo.svg" width="72" height="72" alt="">
  3. </p>
  4. <h1 align="center">Preview Lab</h1>
  5. <p align="center">
  6. <b>Review the UI your coding agent builds, screen by screen, and send the feedback straight back.</b><br>
  7. An <a href="https://agentskills.io">Agent Skill</a> for Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot and 60+ more.
  8. </p>
  9. <p align="center">
  10. <img src="https://img.shields.io/badge/license-MIT-4f46e5" alt="MIT license">
  11. <img src="https://img.shields.io/badge/dependencies-0-4f46e5" alt="Zero dependencies">
  12. <img src="https://img.shields.io/badge/node-%E2%89%A518-4f46e5" alt="Node 18 or newer">
  13. </p>
  14. ![A review round: screens grouped by flow, each with its tips](docs/grid.png)
  15. Your agent captures the screens it changed and opens them in a review page styled like your project. You approve them or mark what to change, right on the screenshot. Press **Send**, and the agent gets short, structured feedback it can act on. Approved screens leave; the rest come back next round, with every change tagged.
  16. ## Quick start
  17. ```bash
  18. npx skills add dbalas/preview-lab
  19. ```
  20. Then ask your agent:
  21. ```
  22. let me review the new checkout screens
  23. ```
  24. That's it. The agent captures the screens, opens the lab in your browser, and waits for you to press **Send**.
  25. ## Why Preview Lab
  26. - **Fast to review.** Screens are grouped by flow. `←` `→` to move, `A` to approve, `C` for changes, drag on a screenshot to mark the exact area.
  27. - **Feedback agents can act on.** Statuses, areas in percent of the screenshot, comments and notes arrive as a few lines of markdown, not a wall of prose.
  28. - **Rounds, not restarts.** Approved screens leave. After a code change only the screens whose pixels changed are tagged **Updated**. Past rounds stay one click away, read-only.
  29. - **Expert tips (beta).** Ask for a *complete review* and the agent adds measured UX, UI, accessibility, responsive and consistency tips, pinned to the screenshot. You tick the ones to apply.
  30. - **Built-in screenshots.** Viewports, dark mode, full pages, single components, and states reached by clicking or typing, using the Chrome, Edge or Chromium you already have.
  31. - **Your brand, not ours.** The lab takes your colors, font, logo and language.
  32. - **Works everywhere.** Local, remote or cloud sessions; light, dark or system theme; desktop or phone.
  33. - **Leaves nothing behind.** Zero dependencies. One lab per checkout, kept in `~/.preview-lab/`, cleaned automatically.
  34. <p>
  35. <img src="docs/detail-dark.png" alt="One screen with tips pinned to it, dark theme" width="72%">
  36. <img src="docs/mobile.png" alt="Reviewing on a phone" width="25%">
  37. </p>
  38. ## Install
  39. | Agent | Command |
  40. |---|---|
  41. | Any of the [60+ supported agents](https://github.com/vercel-labs/skills/blob/main/docs/agents.md) | `npx skills add dbalas/preview-lab` |
  42. | Claude Code (plugin) | `/plugin marketplace add dbalas/preview-lab`, then `/plugin install preview-lab@preview-lab` |
  43. | Gemini CLI (extension) | `gemini extensions install https://github.com/dbalas/preview-lab` |
  44. | Anything else | Copy `skills/preview-lab/` into the agent's skills folder |
  45. The repo also ships an [Agent Plugins](https://agent-plugins.org) manifest (`plugin.json`) for tools that install plugins from a repository. For agents without skills support, add one line to `AGENTS.md`: *"For UI reviews, follow `<path>/preview-lab/SKILL.md`."*
  46. **Requirements:** Node 18+ (22+ for full-page, component and state screenshots), and Chrome, Edge or Chromium for built-in screenshots.
  47. ## Use
  48. Talk to your agent in plain words, or call the skill directly (`/preview-lab` in most agents):
  49. | You say | What happens |
  50. |---|---|
  51. | *let me review the onboarding screens* | Visual review: you give the feedback. |
  52. | *complete review of the checkout* | The agent adds tips for every screen. |
  53. | *only the settings modal, mobile and dark, in Spanish* | Anything you add overrides the defaults. |
  54. ## How a round works
  55. ```
  56. agent you
  57. ───── ───
  58. captures screens ──────────▶ review in the lab
  59. approve · changes · areas · notes
  60. reads the feedback ◀───────── Send
  61. fixes the code
  62. next round ──────────▶ only what changed comes back
  63. ```
  64. The agent receives something like this:
  65. ```markdown
  66. # Preview Lab feedback: shop · round 2 (complete)
  67. 1 approved · 1 changes · 0 unreviewed
  68. **General:** Next round, show me the error state of payment.
  69. **Approved:** Settings
  70. ## Order confirmed [done] — changes
  71. - Area 1 (mobile @ x 8.5%, y 48.7%, 85.3×22.5%): Show item thumbnails
  72. - Apply rec r1 [a11y/high]: The order number is 13px … raise it to 16px.
  73. - Declined recs: r2, r3
  74. ```
  75. ## Remote and cloud sessions
  76. The lab is served on `127.0.0.1:4747` of the machine the agent runs on. If your browser can't reach it, either forward that port, or let the agent run `pl.mjs export`: the round becomes one self-contained HTML file you open anywhere, and **Copy feedback** gives you the text to paste back into the chat.
  77. ## Commands
  78. Your agent runs these; you never have to. They live in `skills/preview-lab/scripts/pl.mjs`.
  79. | Command | What it does |
  80. |---|---|
  81. | `init [--in-project]` | Create or find this checkout's lab. |
  82. | `brand [file]` | Read or set colors, font, logo, language and texts. |
  83. | `add <items.json>` | Check, capture and add items. If anything fails, nothing changes. |
  84. | `recapture [id…]` | Retake stored screenshots. Only changed screens are flagged. |
  85. | `remove <id…>` | Remove items and their screenshots. |
  86. | `serve [--detach]` | Serve the lab and print the feedback when it is sent. |
  87. | `feedback` / `stop` | Print the last feedback, or stop the server. |
  88. | `export [file]` | Save the round as one self-contained HTML file. |
  89. | `round [--keep N]` | Archive the round, drop approved items, start the next one. |
  90. | `contrast <fg> <bg>` | WCAG contrast ratio of two colors. |
  91. | `upgrade` / `clean` / `reset` | Update the lab code, delete unused files, or start over. |
  92. The agent's instructions are in [SKILL.md](skills/preview-lab/SKILL.md), the review guide in [complete-mode.md](skills/preview-lab/references/complete-mode.md), and the lab's design contract in [lab-spec.md](skills/preview-lab/references/lab-spec.md).
  93. ## Troubleshooting
  94. | Problem | Fix |
  95. |---|---|
  96. | No screenshots, `NO_BROWSER` | Install Chrome, Edge or Chromium, or set `PREVIEW_LAB_CHROME` to its path. |
  97. | Full-page or component shots fail | They need Node 22+. Older Node still takes viewport shots. |
  98. | The lab page doesn't open | Open the `PREVIEW_LAB_URL` the agent printed. In a remote session, forward port 4747 or ask for an export. |
  99. | **Send** says the agent is unreachable | The feedback was copied: paste it into the chat. |
  100. | Labs in an odd place | Set `PREVIEW_LAB_HOME`, or ask for the lab inside the project. |
  101. ## Develop
  102. No install step. Run the tests (they also open the lab in a real browser when one is installed):
  103. ```bash
  104. npm test
  105. ```
  106. ## License
  107. MIT · Made with ♥ by [dbalas](https://github.com/dbalas)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.