Communitygithub.com

else-4-pillars-inspiration

Four-pillar inspiration + competitor research that lands as a FigJam board with real embedded screenshots (not just links). Pulls from (1) direct competitors, (2) million-user consumer apps for conventions the audience already knows, (3) the broader ecosystem + aspirational adjacent domains, and (4) visual design trends (Dribbble / Behance / Awwwards). Use this WHENEVER the user wants inspiration, references, a benchmark board, a mood board, competitor research, trend research, or to explore design directions for a screen or flow — including "מחקר השראות", "השראות", "תביא רפרנסים", "בוא נחקור מתחרים", "מה עושים אחרים", "benchmark", "find references", "explore directions" — even if they only name a single source like Mobbin or Dribbble. Always open by asking what the project/topic, audience, and specific screen/pattern are before searching.

What is else-4-pillars-inspiration?

else-4-pillars-inspiration is a Claude Code agent skill that four-pillar inspiration + competitor research that lands as a FigJam board with real embedded screenshots (not just links). Pulls from (1) direct competitors, (2) million-user consumer apps for conventions the audience already knows, (3) the broader ecosystem + aspirational adjacent domains, and (4) visual design trends (Dribbble / Behance / Awwwards). Use this WHENEVER the user wants inspiration, references, a benchmark board, a mood board, competitor research, trend research, or to explore design directions for a screen or flow — including "מחקר השראות", "השראות", "תביא רפרנסים", "בוא נחקור מתחרים", "מה עושים אחרים", "benchmark", "find references", "explore directions" — even if they only name a single source like Mobbin or Dribbble. Always open by asking what the project/topic, audience, and specific screen/pattern are before searching.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/hanigirl/else-4-pillars-inspiration/tree/HEAD/else-4-pillars-inspiration/skills/else-4-pillars-inspiration

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Inspiration Research — The Four Pillars

This skill runs a structured inspiration sweep and lands it as a FigJam board with real embedded screenshots. It exists because "find me some inspiration" usually collapses into a thin list of direct competitors or a pile of Mobbin links. Real inspiration research triangulates from four different angles, each answering a different question — and the designer gets far more out of it when the references are on the board as images they can see, grouped and annotated, not as URLs.

You are the researcher. Your job is to be deliberately broad: a bank app should not only look at other banks. The value is in the angles the user didn't think to ask for.

The four pillars

Each pillar answers a distinct question and has its own primary sources and selection lens. Build one FigJam section per pillar.

#PillarThe question it answersPrimary sources
1Direct competitors"How do products solving my exact problem for my exact audience do this?"Mobbin (search_screens/search_flows), then web search to name the real direct competitors
2Million-user apps"What conventions does my audience already know from the mega-apps they use daily, so I can borrow a familiar mental model?"Mobbin (big consumer apps), web. Think: the patterns billions of people are already trained on
3Ecosystem & aspirational"Who else — not necessarily a competitor — solves a structurally similar problem brilliantly, in an adjacent domain?"Mobbin + web across adjacent domains
4Design trends"What visual / graphic directions are current for this kind of product — where is the aesthetic going?"Dribbble, Behance, Awwwards (via web search + browser), plus Mobbin for shipped-product aesthetics

Pillar 1 — Direct competitors

Same problem, same audience. Start from Mobbin, but also run a web search to make sure you've named the actual direct competitors in this market, not just whoever happens to be in Mobbin. This is the "table stakes" pillar: it tells the user what their audience will expect by default.

Pillar 2 — Million-user apps (familiar conventions)

This is the pillar people forget. The goal is not competitors — it's the giant consumer apps the audience uses every day (Instagram, WhatsApp, Gmail, Spotify, Uber, banking super-apps, Notion, etc.) that have trained the audience on a convention you can reuse. Borrowing a familiar pattern lowers the learning curve to near zero. Pick apps relevant to the specific pattern being designed: researching an onboarding? → how the mega-apps onboard. A feed? → how the feeds the audience already scrolls are built. A number-heavy dashboard? → how consumer finance apps the audience already trusts present numbers. Always tie the pick back to "your audience already knows this from X."

Pillar 3 — Ecosystem & aspirational

Broaden past direct competitors into the surrounding ecosystem and adjacent domains that share a structural challenge. Example — a bank app: pull from fintech, government / public-sector services, big-data and analytics tools, dashboards and any "lots of numbers, high trust, high stakes" product. The lens is structural similarity, not category. These are where the genuinely fresh, best-in-class ideas usually come from.

Pillar 4 — Design trends

Shift from product flows to visual / graphic direction. Surface ~10 current visual directions for this kind of product so the user gets a read on where the aesthetic is going (type, color, density, motion cues, illustration style, graphic language). Dribbble is the anchor source; add Behance / Awwwards for richer, more produced directions, and Mobbin for how the trend looks once it actually ships. This pillar is inspiration, not a spec — label each as a direction, and say what's borrowable.

Step 0 — Always interview first

Before any search, ask the user (even if the project folder seems to tell you — the user asked to always be asked). Ask through the interactive question UI (the structured multiple-choice question card — AskUserQuestion), not as a plain-text list. Give concrete options for each (with example screens/platforms/audiences so the user just taps) and let them type anything else via the free-text option. This is a hard requirement — the user asked for the questions to come through that UI.

Ask the four choice questions in one interactive card (project, audience, screen, platform). The two inputs below that can't be multiple-choice — user research and the board URL — come as the card's free-text option or a one-line follow-up:

  1. Project / topic — what is being designed, and what does the product do?
  2. Audience / persona — who uses it; their sophistication and what apps they already live in (this directly feeds pillar 2).
  3. The specific screen / pattern / flow being researched — onboarding, a dashboard, an AI-generation flow, a checkout. Be concrete; vague briefs produce vague boards.
  4. Platform — mobile / web / both (so Mobbin platform is set correctly).
  5. User research — the engine of "why this is good for us." Ask, in words close to these: "מה כבר ידוע לנו על המשתמשים? מה הם מחפשים, מה הכאבים שלהם, איך נראה מסע המשתמש שלהם?" Invite the user to paste one or more links (a research doc, interview notes, a journey map, a PRD) or to write it as free text. If links are given, open and read them (built-in browser / web-fetch; for a Figma/FigJam/Notion/Docs link use the matching tool, not a raw fetch) and pull out the concrete pains, needs, and journey moments. This is optional but high-leverage — see the note just below.
  6. Where to place the board — the Figma / FigJam file URL (or create a new file). A URL is free-text, so ask it as the free-text option or a one-line follow-up.

If the user already gave some of this in their request, confirm the gaps rather than re-asking everything. Do not start searching until the pattern is concrete.

Why the research question matters most

"מה רואים" (what you see) is objective and easy. "למה טוב לנו" (why it's good for us) is where the research lives — and it's only valuable when it's tied to a real user pain, need, or journey moment, not generic praise ("nice clean layout"). The research from question 5 is the lens: every reference you keep should earn its place by answering "which known pain / need / journey stage does this help solve?" Distil what the user gave you into a short working list of pains → needs → journey stages, and carry it into Step 2 (selection) and Step 3 (the caption).

If the user has no research to share, say so plainly: the "why for us" lines will be assumption-based (grounded in the persona, flagged as a hypothesis), and offer to sharpen them later once real research exists. Don't fake confidence the research doesn't support.

Step 1 — Research each pillar

Default target: 5 references per pillar (≈20 total). You have a standing license to go beyond 5 for any pillar when you see real research value — e.g. the trend pillar splits into several distinct visual directions worth showing, or a pillar surfaces a cluster of ideas the user would clearly benefit from. When you expand, say why in one line. Don't pad to hit a number.

For each pillar:

  • Mobbin first (mcp__mobbin__search_screens, platform: 'web' or 'ios' for mobile-first, mode: 'deep', limit: 6-8). Run several angled queries in parallel. Vary the query per pillar — pillar 2 queries name the mega-apps; pillar 3 queries name the adjacent domains.
  • Web search to (a) name the real players a database might miss, and (b) reach sources with no MCP. Always web-search pillar 1 (confirm the true direct competitors) and pillar 4 (current Dribbble/Behance/Awwwards directions).
  • Dribbble / Behance / Awwwards have no MCP here. Use web search to find specific shots/pages, then open them in the browser to capture or download a real image. If a site blocks image embedding, fall back to a browser screenshot of the shot, or — last resort — a link-only card, and say so.
  • Flows, not only screens. mcp__mobbin__search_flows returns a real multi-screen flow (a whole sequence), not a single still. Use it to find the one winning flow (see Step 3), and any time a sequence teaches more than a single frame.

Flows-only mode: if the user explicitly asks for flows (not "references" / "screens" / "inspiration"), the entire output is flows — use search_flows for every pillar and embed the multi-screen sequences, never single shots. Otherwise the board is single-screen references plus the one winning flow (Step 3).

Keep for every pick: the product/source name, a one-line hook, the real screenshot image URL (Mobbin short URL, or a downloaded image path), and the source link.

Step 2 — Select the strongest

Apply benchmark criteria (from the ux-strategy:competitive-analysis methodology):

  1. Pain / journey fit — does it help solve a known pain, need, or journey moment from the Step 0 research? This is the primary lens: a beautiful reference that answers no real pain doesn't earn a slot. (When there's no research, this softens to persona fit, flagged as a hypothesis.)
  2. Pattern fit — does it map directly to the screen/flow in the brief?
  3. Persona match — would this audience actually use / recognize it? (Decisive for pillar 2.)
  4. Aesthetic match — does the visual register fit what's being built? (Decisive for pillar 4.)
  5. Novelty vs. what's already on the board — don't repeat a pattern another pick already shows.

Mark the single best pick in each pillar as ⭐ anchor.

Step 3 — Build the FigJam board

Use the figma-inspiration-board skill for the screenshot mechanics only — download each shot locally, upload_assets in batches of 5 (max count per call), and set each image as a FILL fill on a figma.createFrame(). That skill's card + sticky + connector layout is not the target here. Use the item anatomy below instead — it matches the reference inspiration board the user works from.

Item anatomy — the required format

Every reference on the board is a vertical stack of three elements, nothing more:

┌─────────────────────────────────────────┐
│  ⭐ ProductName — what it is             │  ← TITLE (text, bold ~Inter 72, full column width)
│                                         │     ⭐ prefix only on the pillar's anchor pick
└─────────────────────────────────────────┘

┌─────────────────────────────────────────┐
│                                         │
│          [ REAL SCREENSHOT ]            │  ← IMAGE FRAME (createFrame + image FILL)
│                                         │     portrait for mobile, landscape for web
└─────────────────────────────────────────┘

┌─────────────────────────────────────────┐
│  מה רואים: <one-line description of the  │  ← CAPTION (single text note under the image)
│  screen / pattern / visual>.            │
│  למה טוב לנו: <why it helps OUR product, │
│  tied to the persona + the screen being │
│  designed>.                             │
│  🔗 <source link>                        │
└─────────────────────────────────────────┘

So each item = title → screenshot → caption. No rounded-rect cards, no connector arrows. The caption is one text block, and it always carries the two things the user asks for every time:

  1. מה רואים / What you see — a plain, factual read of what's in the shot (the layout, the pattern, the visual move).
  2. למה זה טוב למוצר שלנו / Why it's good for our product — the payoff, phrased against a specific user pain, need, or journey moment from the Step 0 research (name it: "פותר את הכאב ש…" / "עונה על הצורך ל…" / "בשלב X במסע"). Anchor it to the real project, persona and screen — not generic praise; a reference that maps to no known pain is just a pretty picture. If there was no research, write this as an explicit hypothesis ("הנחה: אם המשתמש מתקשה ב…") rather than stated fact. Sharpen the line per pillar:
    • Pillar 1 — is this a table-stakes expectation to meet, or a gap to beat?
    • Pillar 2 — name the mega-app: "האודיינס כבר מכיר את זה מ-" / "your audience already knows this from ."
    • Pillar 3 — the transferable structural idea ported from the adjacent domain.
    • Pillar 4 — the visual direction and which elements are borrowable (type / color / density / motion).

Write the caption in the user's language (Hebrew for this user), matching the tone of their existing board.

The one winning flow (⭐ — a full sequence, not a single shot)

A single screen shows a moment; a flow shows the thinking — how a product actually walks the user through the job, step by step. On every board, pick exactly one standout flow (the one you'd genuinely tell the user to go study) and give it the spotlight:

  • Pull the actual multi-screen flow with mcp__mobbin__search_flows — not a single search_screens shot. Download and embed its screens in sequence, left → right, so the row reads as steps 1 → 2 → 3 → ….
  • Give it its own strip, set apart from the four pillars — a section titled ⭐ Winning flow — <Product>: <flow name>, placed first on the board (above the pillars) so it reads as the headline, not buried in a pillar.
  • Caption it once, under the strip, with the approach (what the flow does, step by step, in a line or two) and why it fits — tied to the Step 0 pain / journey, exactly like every other caption.
  • Earn it, don't force it: choose the single best flow on merit, and if nothing clearly stands out, skip the winning flow entirely rather than inventing one.

(In flows-only mode this isn't "one" — every pillar is built from flows. The winning-flow strip is for the normal single-screen board, where it's the one sequence worth the extra space.)

Board structure & layout

Four sections, one per pillar, each clearly titled, so the four read as a set (give each a distinct fill tint):

  • פילר 1 — מתחרים ישירים / Direct competitors
  • פילר 2 — אפליקציות ענק (קונבנציות מוכרות) / Million-user apps
  • פילר 3 — אקוסיסטם והשראות / Ecosystem & aspirational
  • פילר 4 — טרנדים עיצוביים / Design trends

Add a short header note at the top of each section stating the question that pillar answers, so the board explains itself to anyone who opens it later.

Lay items as columns in a row, left → right, with the three elements stacked per column. Working coordinates that match the reference (section-local): columns spaced ~2800px apart (title/image/caption ~2280px wide); within a column, title at y≈980, image frame at y≈1250 (~1700 tall for portrait mobile), caption at y≈3100. Five items per row; if a pillar earns more than 5, start a second row below (repeat the y-offsets lower down). Mark the pillar's best pick with a ⭐ on its title.

Step 4 — Report back

When the board is built, reply with:

  • The FigJam board link.
  • A short top-picks summary: the ⭐ anchor from each pillar and the single sharpest takeaway, so the user gets the headline without opening the board.
  • Any gaps or fallbacks — a pillar where Mobbin was thin, a trend shot that could only be linked, a source that blocked embedding.

Step 5 — Offer another round

A board is a starting point, not a finish line — the user often wants to go deeper on one angle. After reporting, ask through the interactive question UI (AskUserQuestion) whether to pull more directions / inspirations, and let them pick which pillar(s) — set multiSelect: true with the four pillars as options plus an "כל הפילרים / All pillars" option. (If they want more only on, say, pillar 4, they tick just that.)

For each chosen pillar, run another round:

  • Search fresh angles you haven't used — new Mobbin queries, new adjacent domains, different Dribbble directions. Don't repeat a product or pattern already on the board; the whole point is new ground.
  • Apply the same selection lens (pain/journey fit first), download, upload, and append to that pillar's existing section as a new row below the current one — reuse the row y-offsets, shifted down by one row-height (see the layout note: "if a pillar earns more than 5, start a second row below"). Grow the section's height to fit.
  • Keep the ⭐ anchor as-is unless a new pick genuinely beats it — then move the star and say why.

Then offer again. Repeat until the user says they're done. Each round is cheap relative to its value, so keep the loop open rather than assuming the first pass was enough.

Edge cases & fallbacks

  • Mobbin returns little for a pillar — widen to web search and name real products, then pull their shots via browser. Don't leave a pillar empty; say if a pillar is genuinely sparse and why.
  • Dribbble / Behance blocks the image — browser-screenshot the shot, or fall back to a titled link card. Never silently drop it.
  • No Figma file given — ask; create a new FigJam file only once the user confirms where.
  • Scope too broad ("inspiration for my app") — push back to Step 0 and pin down one screen/pattern. Four pillars across a vague brief produces four vague sections.
  • Mobile vs web — match Mobbin platform to what's being designed; a mobile app should pull ios shots, not web.

Related Skills