ui visual consistency checker

    1

    Audit UI consistency across screenshots and code to detect design drift, normalize tokens, and ensure visual hierarchy.

    Free

    2 usesSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    ui visual consistency checker

    Example session with this skill installed

    Review this CSS snippet and screenshot for visual consistency. Focus on button padding and typography roles. Report any drift from a standard 4px grid.

    • Read your context and instructions
    • Compiled the ui visual consistency

    I have completed a specialized review of the provided CSS and screenshot, focusing on spacing rhythm and typography hierarchy.

    Visual Consistency Audit

    Finding IDCategoryEvidenceObservationRiskRecommendation
    VC-001SpacingCSS button.primarypadding: 10px 18px;OBSERVED 2px drift from 4px grid (Expected 8/12/16/20).Normalize to padding: 8px 16px;.
    VC-002SpacingCSS button.icongap: 7px;OBSERVED Odd-pixel spacing breaks vertical alignment rhythm.Update to gap: 8px; (2u).
    VC-003TypeScreenshot"Submit" label vs "Cancel"CALCULATED Weight mismatch. Primary is 600, Secondary is 400.INFERRED Intentional hierarchy for prominence.
    VC-004ColorCSS .btn-errorbackground: #e02424;OBSERVED Hex code is hardcoded; no variable used.RECOMMENDATION Map to --color-status-error.

    Token Inventory (Inferred)

    • Space-Unit: 4px
    • Typography-Primary: Inter, sans-serif
    • Color-Action: #2563eb (Blue 600)
    • Color-Surface: #ffffff

    Intentional vs Accidental Assessment

    The 10px/18px padding in VC-001 appears to be accidental drift, as it does not align with the 8px or 12px patterns found in the navigation elements. The weight difference in VC-003 is categorized as an intentional hierarchy variant to distinguish primary actions from dismissive actions.

    Next steps

    1. Review the proposed --color-status-error mapping in your global CSS variables.
    2. Update the button padding to align with the 4px grid system.
    3. Verify if the 7px icon gap was a specific optical adjustment or a typo.

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

    About this skill

    The problem

    UI drift happens when multiple developers build features without strict token adherence, leading to fragmented spacing, inconsistent typography, and "magic" hex codes. Manual visual QA is slow, subjective, and often misses subtle variations in component states or alignment logic.

    What it does

    • Analyzes screenshots, HTML, and CSS to inventory visual primitives like colors, font scales, and spacing intervals.
    • Distinguishes between intentional hierarchy and accidental inconsistency across components.
    • Clusters typography and color roles to infer candidate design tokens for normalization.
    • Generates prioritized visual QA checklists based on evidence confidence and implementation effort.
    • Audits alignment, container logic, and component anatomy for systemic drift.

    Why this beats prompting it yourself

    Generic prompts often hallucinate design facts or recommend "fixing" intentional state changes, like hover or disabled styles. This skill uses a structured decision model to separate raw primitives from semantic roles, ensuring recommendations don't break functional hierarchy or accessibility.

    Use cases

    • Reviewing a new feature implementation against existing design system references.
    • Auditing a legacy codebase to extract a baseline set of design tokens.
    • Performing a visual regression check after a major CSS refactor or framework migration.
    • Preparing a developer handoff audit to ensure component consistency before a release.

    Known limitations

    Cannot prove runtime interaction behavior, browser-specific rendering bugs, or screen-reader accessibility from static screenshots alone. Requires clear source-of-truth constraints for high-confidence audits.

    Cannot prove runtime interactions or browser rendering bugs from static assets. Requires design references or source-of-truth constraints to move beyond INFERRED status.

    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

    2 uses

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Permissions

    No special permissions declared or detected

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    The skill accepts static screenshots (PNG/JPG) alongside HTML and CSS files to analyze the relationship between your code and the rendered output for precise drift detection.

    Unlike basic visual regression tools, this skill uses a structured model to distinguish between intentional design choices (like hierarchy or state changes) and accidental 'magic' values that need normalization.

    This skill is designed to work with any LLM-based agent capable of processing images and code snippets, such as those built on GPT-4o or Claude 3.5 Sonnet.

    You receive the optimized agent instructions, a structured analysis framework for token extraction, and a guide on how to provide 'source of truth' constraints for the best results.

    Yes, it is specifically designed to cluster fragmented CSS values into a unified set of candidate design tokens, making it an ideal first step for migrating to a formal design system.

    New in Frontend & Web Apps