Figma System-First Builder: tokens first, auto-layout always, verified by measuring

    1

    Build Figma files that stay editable: variables and text styles first, auto-layout from the first frame, components for repeats, checks by measuring.

    $9

    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 CLIDeepSeekDeepSeekMuseMuse+22 more

    Figma System-First Builder: tokens first, auto-layout always, verified by measuring

    Example session with this skill installed

    Create a new Figma file for a SaaS landing page: desktop (1440px), tablet (768px), mobile (375px). Need Light/Dark modes. Components: header, hero, feature grid (3 cols), testimonial carousel, footer. Text styles for each breakpoint. Build system first, then first frame.

    • Read your context and instructions
    • Compiled the figma system-first builder

    The skill runs: Phase 0 — asks for design system origin (reuse/extract/guided/autonomous). Phase 1 — creates Tokens collection with 8 color variables (ink, text-muted, accent, bg, surface, rule, white, accent-soft) and 2 modes (Light/Dark); creates Tokens/Spacing numeric variables (4,8,12,16,24,32,48,64,80,96). Phase 2 — creates text styles WEB/Display/Hero, WEB/Title/Section, WEB/Text/Body, WEB/Text/Small, WEB/Data/Metric, WEB/Chrome/Nav for each breakpoint set (WEB-DESKTOP, WEB-TABLET, WEB-MOBILE) with same role names. Phase 3 — draws first frame (page auto-layout vertical, header + hero + features + testimonials + footer in flow, fills=[]). Phase 4 — converts header, feature card, testimonial card, footer into components with text properties; feature grid as component with Slide=01/02/03 variants. Phase 5 — reads back every node, verifies measurements match tokens, no 100px cross-axis, no absolute positioning, no fills on containers.

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

    About this skill

    Most Figma files rot because they start with rectangles and retrofit structure later — resulting in forty slightly different grays, six text styles that should be one, and no way to change a color without hunting through hundreds of frames. The method comes from a real project: a portfolio built in four versions (web, A4 print, slides, CV) inside one Figma file, with 9 color variables and 8 components, all in auto-layout and still editable without breaking.

    The method enforces a strict build order: 0) clarify design system origin (reuse existing, extract from reference, guided questions, or autonomous generation); 1) create 6–9 semantic color variables (ink, text-muted, accent, bg, surface, rule, white, accent-soft) using Figma Variables with native modes for Light/Dark; 2) define spacing scale as numeric variables or closed set (standard 4/8pt base: 4, 8, 12, 16, 24, 32, 48, 64, 80, 96) — no off-scale values; 3) create text styles with SET/Group/Name taxonomy (e.g., WEB/Text/Body, SLIDE/Title/Section) — same role names across supports, 10–14 styles per support; 4) draw first frame using only those tokens; 5) convert anything appearing 3+ times into a component with text properties for content variation and variants for state/position; 6) verify by reading actual node measurements (dimensions, gaps, padding), not screenshots — catch 100px fixed-axis frames, undeclared absolute positioning, fill leaks.

    The skill drives Figma via the official MCP connector (use_figma tool) and works alongside the official figma-use skill (which handles API mechanics). It includes battle-tested recipes for: creating frames that don't explode to 100px on cross-axis; stripping fills from auto-layout containers; handling SPACE_BETWEEN with fill children; loading fonts before writing text; fixing image aspect ratios in auto-layout; reading pages sequentially; counting component instances correctly; converting "one component per text" anti-patterns into single components with text properties; attaching existing colors to variables; assigning text styles to legacy files by exact weight match.

    What distinguishes it: system-first discipline prevents structural debt; auto-layout from frame one avoids conversion nightmares; component properties over component-per-text scales to bilingual/multi-page projects; verification by measurement catches API quirks (100px fixed cross-axis, white fills, collapsed gaps) that screenshots miss; token economy — load only the reference file needed for the current step (auto-layout for nodes, system-first for tokens, api-recipes for code, naming-and-hygiene for cleanup).

    The package includes: SKILL.md with complete phased method and context hygiene rules; references/system-first.md (color variables, spacing tokens, text style taxonomy, visual lexicon); references/api-recipes.md (10 traps + 8 recipes with tested code snippets for use_figma); references/auto-layout.md (container rules, page structure, section blocks, image-caption cells, absolute positioning isolation); references/naming-and-hygiene.md (layer naming convention, file handoff checklist); references/claude-design.md (publishing reusable system cards).

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Workflow Automation.

    ["Builds inside Figma through the official MCP: it makes design files, it does not write front-end code","Needs write access to the file; with view-only access it can only audit","Version-history save points must be created by hand in Figma","Brand choices (type, palette, measures) come from you or from a few guided questions; there is no template file","Large files are read one page at a time, so a full audit takes a few passes"]

    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 buying a skill.

    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 3 days ago

    Accounts
    Figma
    Software
    Official skill "figma-use" (recommended, loaded first)

    Permissions

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    Yes, recommended. `figma-use` handles API mechanics (tool calls, authentication, node types). This skill provides the method — what to build and in what order. Load `figma-use` first, then this skill.

    Yes. The recipes cover: attaching existing fills to color variables (exact match), assigning text styles by exact weight + closest size/leading, converting "one component per text" into single components with text properties using `swapComponent` + `setProperties`. It reads by node name, not ID, so it works across pages.

    The method allows two only with justification. One typeface with 3–4 weights is standard. Two is a conscious choice documented in the system origin (Phase 0).

    Use Figma Variable modes inside the same `Tokens` collection. Define `color/ink`, `color/bg`, `color/surface` etc. once, then add `Mode = Light` and `Mode = Dark` values. No file duplication, no frame duplication.

    Yes. The method applies to Figma, FigJam, and Figma Slides — any surface where you create frames, variables, styles, and components. Support-specific style sets (e.g., `SLIDE/Text/Body`) handle differing body sizes.

    `figma.createFrame()` defaults to 100×100 FIXED/FIXED. Setting `layoutMode` only makes primary axis HUG. The skill mandates: after `layoutMode`, explicitly set `layoutSizingHorizontal = 'HUG'` AND `layoutSizingVertical = 'HUG'` on every new frame. This is in `references/api-recipes.md` and enforced in the workflow.

    Phase 5: after each block, read the node and check actual measurements — dimensions, itemSpacing, paddingLeft/Right/Top/Bottom. Verify no frames stuck at 100px on cross-axis. Verify no undeclared absolute positioning. Verify container fills are `[]`. Screenshots lie; measurements don't.

    No. The method governs build order and rigor; your brief decides fonts, measures and palette.

    New in Workflow Automation