Design QA: Reference Capture

    1

    Deterministic, comparable screenshots of any page or mockup per viewport, including states, folds, elements and a provenance record.

    $7

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Design QA: Reference Capture

    Example session with this skill installed

    Capture https://example.com/pricing as the reference for iOS and desktop. I also need the hover state of the primary button and the pricing table as its own image. Hide the cookie banner.

    • Read your context and instructions
    • Compiled the design qa
    • Generated the UI component

    Source classified as

    open (public page), so I captured it directly. Settings: DPR 1, animations frozen, reduced motion, locale en-US, timezone UTC, waited for fonts.

    Commands

    python3 S/scripts/snapshot.py --url https://example.com/pricing --name ref --out .design-qa/pricing/ref --vp ios,web --dom --axe --license open --block-external
    python3 S/scripts/capture_states.py https://example.com/pricing --name ref --out .design-qa/pricing/ref --vp ios,web --hover ".btn-primary" --element ".pricing-table" --hide ".cookie-banner"
    

    Files written (.design-qa/pricing/ref/)

    Provenance (ref.source.json, excerpt)

    {"source": "https://example.com/pricing", "method": "playwright-chromium",
     "license": "open", "viewports": ["390x844", "1440x900"],
     "options": {"dpr": 1, "block_external": true, "hide": [".cookie-banner"]}}
    

    Notes for the comparison step

    • axe: 1 contrast issue on desktop (footer links), already present in the reference
    • Suggested mask for the live chat bubble on iOS: 390x844:326,780,390,844
    • Capture the candidate with the same viewports and the same --hover / --element selectors so states compare like with like.

    390x844.png

    TSX · React component

    Generated

    Example file from a real run - the skill writes it into your workspace.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Visual comparisons are only fair when both sides are captured the same way. This skill gives your agent a fixed capture recipe: device scale 1, frozen animations, reduced motion, fixed locale and timezone, and capture only after fonts are ready.

    What it captures

    • Exact viewports: iOS 390x844, Android 412x915, desktop 1440x900 or any WxH
    • First fold, later folds, hover and keyboard-focus states and single elements
    • Paints over cookie banners and other selectors you name
    • Computed design tokens (colours, fonts, sizes, spacing, radii) and axe-core contrast
    • 2x or 3x phone screenshots converted to a reference in CSS pixels. Screenshots from the wrong device class are refused instead of stretched
    • A recipe for PDF mockups and Figma exports

    Provenance and usage classes

    Every capture writes a provenance file with source, time, machine, usage class and options. Usage classes guide what you may capture: open pages are captured, login pages need a user-supplied screenshot, restricted pages are linked only. The skill never logs in or crawls.

    The output feeds straight into the Design QA Fidelity Check.

    What you get

    • SKILL.md: an 8-step capture workflow with usage-class rules
    • scripts/snapshot.py: first-fold or full-page capture with DOM tokens, axe and provenance
    • scripts/capture_states.py: folds, hover, focus, element screenshots, hide selectors
    • scripts/prep_ref.py: phone screenshot to a reference at exact CSS size
    • references/capture-guide.md: frozen settings, viewport presets, masks, PDF and fonts rules, library layout

    Permissions / network

    • Visits only the URLs and files you provide (plus their own sub-resources). --block-external keeps it local or same-origin
    • Never logs in or crawls. No telemetry, no uploads
    • Writes only to the output folder you choose (default .design-qa/)

    Web fonts, consent banners and lazy-loaded content on real sites need --hide, --wait load or masks. Captures are comparable only with candidates rendered on the same machine. Downscaled phone screenshots are slightly softer than native 1x renders.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 day ago

    Requires

    Python 3.10+
    Playwright with Chromium
    Python

    Permissions

    Terminal / Shell
    Browser
    Read Files
    Write Files
    Network Access

    File Scopes

    .design-qa/**

    Terminal runs the bundled Python scripts. Browser/Network: headless Playwright Chromium visits only the URLs and files you provide (and their own sub-resources); --block-external keeps it local or same-origin. Never logs in or crawls. Install contacts PyPI, the Playwright CDN and optionally npm. Writes only to the output folder (default .design-qa/).

    Listed1 day ago

    What's inside

    Frequently Asked Questions

    No, by design it never logs in or crawls. For logged-in pages you provide a screenshot and the skill converts it into an exact-size reference.

    If a screenshot is not an integer scale of the chosen viewport (for example 1080x2400 is a 360x800 screen), it is refused instead of stretched. Use the matching viewport for both sides or capture again.

    No telemetry and no uploads. The browser only loads the page you give it and that page's own resources. Use --block-external to keep third-party requests out.

    Yes. Export the Figma frame at 1x and the exact viewport size, or rasterise a PDF page with pdftoppm using the documented recipe.

    New in Testing & Debugging