HTML Import

Use this skill when the deliverable is editable Figma layers generated from HTML, a local .html file, or a simple web URL.

FigCraft's native path converts HTML into a create_frame payload so the result stays inside the declarative pipeline: Opinion Engine, design preflight, lint, verification debt, and token/component workflows still apply.

Skill Boundaries

• Use import_html for static HTML, local HTML files, and simple server-rendered pages.
• Local filePath imports require .html, .htm, or .xhtml; by default they are limited to the MCP server cwd. Set FIGCRAFT_HTML_IMPORT_ROOTS to allow additional roots.
• URL imports are for public http(s) pages. Private/loopback URLs such as localhost are blocked by default; set FIGCRAFT_HTML_IMPORT_ALLOW_PRIVATE_URLS=true only when intentionally importing a local/private page.
• Use dryRun:true first for unfamiliar HTML to inspect the generated create_frame payload.
• If the source is a live web app with JavaScript rendering, external CSS, animations, or complex responsive behavior, use the official Figma MCP generate_figma_design as a pixel reference when available, then rebuild/refine through FigCraft tools.
• If the task is ordinary from-scratch UI creation without an HTML source, use figma-create-ui.
• If the task is mapping code components to Figma components, use figcraft-code-connect; Code Connect's figma.html is not an HTML import path.

Workflow

1. Call get_mode first. If it fails, stop and ask the user to open Figma -> Plugins -> FigCraft.
2. Load ui-ux-fundamentals, plus design-guardian when a library is selected or design-creator when no library exists.
3. Inspect the source:

• Inline snippet: pass html.
• Local file: pass absolute filePath.
• Static URL: pass url; JavaScript is not executed.

1. For non-trivial HTML, call import_html({ ..., dryRun:true }) and inspect:

• root dimensions and name
• generated children
• unresolved image warnings
• node count and depth truncation warnings

1. Write with import_html only after the payload looks reasonable.
2. Verify the result:

• export_image on the returned root frame
• lint_fix_all
• verify_design if _qualityScore < 80, _qualityWarning exists, or _verificationDebt remains

Native Converter Contract

import_html is conservative by design.

It supports:

• semantic containers: main, section, article, header, footer, nav, div, form
• text: headings, paragraphs, spans, labels, links, buttons
• controls: input, textarea, select
• media: remote img URLs and inline svg
• inline CSS basics: color, background color, border, radius, padding, margin, gap, width, height, font size, font weight, line height, text decoration, flex direction

It does not currently support:

• JavaScript execution
• full external CSS cascade
• CSS Grid fidelity
• pseudo-elements
• canvas/video rendering
• local relative images as image bytes
• exact browser layout parity

Follow-Up Fixes

After import, prefer normal FigCraft edits:

• Use nodes(method:"update") for layout, fills, strokes, sizing, and text props.
• Use text(method:"set_content") or text(method:"set_range") for content changes.
• In library mode, replace hardcoded colors/text styles with search_design_system + variables_ep(method:"batch_bind") where appropriate.
• For repeated UI pieces discovered in the import, convert stable patterns into components with create_component_from_node or rebuild them via create_component.

Examples

Dry-run a local file:

{
  "filePath": "/absolute/path/page.html",
  "name": "Imported / Pricing Page",
  "width": 1440,
  "dryRun": true
}

Import an inline snippet:

{
  "html": "<main><h1>Launch faster</h1><p>Design and ship with one system.</p><button>Start</button></main>",
  "name": "Imported / Landing",
  "width": 1200
}