mikefutia

Claude Higgsfield Figma — AI skill for Claude Code

AI community

Claude Code skill: generate text-free AI image plates in Higgsfield, then build a full Meta ad variant grid in Figma where headline, body, CTA, logo and brand colors stay live editable layers.

How to install Claude Higgsfield Figma

This entry records only its repository, not the path inside it, so there is no exact command to give. Open mikefutia/claude-higgsfield-figma and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Claude Higgsfield Figma does

Claude Code skill: generate text-free AI image plates in Higgsfield, then build a full Meta ad variant grid in Figma where headline, body, CTA, logo and brand colors stay live editable layers. By Mike Futia — SCALE AI.

Alternatives in AI

Claude Vision — Claude Vision Skill (Mike Futia SCALE AI) 97 ★
Tmux Agent Indicator — Tmux plugin that gives visual feedback for AI agent states (running/needs-input/done) 91 ★
Campfire — Campfire - Collaborative web platform for AI coding agents with Collective Intelligence layers 57 ★

README

Higgsfield → Figma

A [Claude Code](https://claude.com/claude-code) skill that produces Meta ad variants where every element a designer might change — headline, body, CTA, logo, brand colors — stays a live editable layer. Only the imagery comes from AI.

The reframe

The request usually arrives as *"make this PNG editable."* That is not possible — the text is pixels. Auto-trace, OCR-and-retype, background removal are all recreation with extra steps, which is what the designers are already doing by hand.

The fix is upstream: **stop generating finished ads.** Generate only the image *plate* — scene, product, lighting, no type — and let Figma own everything else. That is a prompting change before it is a tooling change.

Once plates live in a variables-and-components file, the small edits that started this stop being recreation work:

Edit	Before	After
Change a headline	rebuild the ad	click, type
Change button color across every variant	rebuild every ad	change one token
Swap the image	rebuild the ad	swap one fill
Light version of the whole set	rebuild every ad	one mode switch

That is the actual deliverable — not the ads, the fact that changing them is cheap.

What it does

Runs end to end, no other skill required:

Reads your Figma templates to find each placement's protected zone
Writes copy from an angle brief (speed-of-results, price-objection, …), not finished copy
Generates plates via Higgsfield — natively per ratio, never cropped across ratios
QA's every plate and rejects on stray lettering, protected-zone intrusion, or wrong luminance
Uploads to Figma and normalizes to exact placement sizes
Builds the variant grid — clone template, set plate fill, switch one variable mode, set copy
Names frames so Figma builds the export folder structure for you

Default placements are the Meta set: `4x5` (1080×1350), `1x1` (1080×1080), `9x16` (1080×1920).