| name | higgsfield-assets |
| description | Use whenever a visual or media ASSET is needed for an app or page — an image, icon, logo, illustration, hero/background, avatar, texture, video clip, sound effect, or voiceover. Higgsfield is the ONE path for generating assets: never use stock photos, emoji, placeholder boxes, hand-drawn SVGs, or another generator, and never ask the user to go find assets. AUTO-TRIGGER on asset intent: 'add an image/icon/logo/illustration', 'needs a hero', 'a background for…', 'generate a picture/video/sound', or any build step that requires visual/media content. Scope is assets ONLY — do NOT use this for code, data, schema, copywriting, or UI layout. |
| metadata | {"author":"dev@potenlab.dev","version":"0.1.0","date":"2026-06-15T00:00:00.000Z","status":"v1","reuses":"higgsfield MCP, verify-in-browser"} |
higgsfield-assets — every asset is generated with Higgsfield, automatically
Think as a developer, respond as a designer/PM. You pick presets, prompt, and place assets like
an engineer wiring real files; you report like a PM ("I generated a clean hero image and the icons —
they're in and look right on phone and desktop").
The rule (non-negotiable)
Any visual or media asset a build needs is generated with the Higgsfield MCP — every time, without
being asked. The moment a UI needs an image, icon, logo, illustration, hero, background, avatar,
texture, video, sound effect, or voiceover, that is a Higgsfield job. This is automatic, not a step the
user has to request.
Do not, for an asset:
- use stock photos, clip-art, emoji, Unicode glyphs, or
<svg> you hand-author as a substitute,
- drop in a grey placeholder box /
placeholder.com / picsum / lorem-image URL,
- generate the asset with any other image/video/audio tool, or
- tell the user to "find an image" or "drop in a logo" themselves.
If you are about to ship a placeholder or ask the user to source an asset — stop and generate it with
Higgsfield instead.
Scope — assets ONLY
This skill is only about generating visual/media content. It does not apply to, and must not
expand into:
- application code, wiring, or routing,
- the data model, schema, RLS, or migrations (that's
supabase-integration / supabase),
- copywriting / body text / UX copy, or
- UI layout and component structure (that's
frontend-design / impeccable).
Higgsfield is the asset engine, not the app builder. Stay in your lane: pictures, motion, and sound.
Access is already gated
The plugin verifies Higgsfield is connected and signed in before any work starts (the
using-designer-workflow hard gate). Do not re-run the full gate here — just generate. If a generation
call returns an auth/connection error mid-build, surface the same /mcp → higgsfield → Authenticate
sign-in instruction and pause.
How to generate (pick the right tool)
Higgsfield MCP tools are deferred — load them with ToolSearch (select:mcp__...higgsfield...) the
first time you need them. Then choose by asset type:
| Need | Higgsfield tool |
|---|
| Image, icon, logo, illustration, hero, background, avatar, texture | generate_image |
| Video / motion clip | generate_video |
| Sound effect, music bed, or voiceover | generate_audio |
| 3D asset | generate_3d |
| Sharpen / enlarge an existing asset | upscale_image, upscale_video |
| Transparent overlay (cut out the subject) | remove_background |
| Extend or re-aspect an image for a layout | outpaint_image, reframe |
Helpers: presets_show / models_explore to choose a brand-locked preset/model before generating
so assets stay on-brand and consistent across the app. When the user provides their own local
photo/logo to work from, call media_upload_widget; for a web URL, call media_import_url first and
pass the returned media_id (never paste a raw URL into a generation param).
The loop
- Spot the asset need — as you build or restyle, list every image/icon/media slot the UI has.
- Pick a preset (
presets_show) so the set is brand-consistent, then generate each asset at
the right dimensions/aspect for where it lands (hero vs. icon vs. avatar).
- Place them into the app yourself — wire the real generated files into the components. Don't leave
a slot empty or a placeholder behind.
- Verify — when the assets are in, the change is "runnable", so
verify-in-browser applies:
confirm they actually render at mobile + desktop and nothing is broken/blurry/missing.
Report back (designer/PM language)
Translate — never surface tool names, prompts, or asset IDs in chat:
"I generated the artwork with our brand presets — a hero image, the nav icons, and a little empty-state
illustration — and placed them in. They render cleanly on phone and desktop."
The raw prompts/preset choices, if worth noting, go in the PR, not the chat.
