agentleFS
jeremyinthebayjeremyinthebay/relay-skills/skills/brand-logos-without-rehosting/SKILL.md

brand-logos-without-rehosting

Skill1 starsChanged 4 months agoMITView on GitHub

Show real company logos next to brand names — airlines, hotels, banks, SaaS vendors, anything — without scraping or rehosting a single asset file. Resolves each mark at runtime from the brand's own published favicon, with a self-hosted override and a monogram fallback so a missing logo is never a broken image. Use when building a comparison table, directory, vendor dashboard, or any list where a logo makes a row scannable. Also use when logos look fuzzy, when a logo resolves to the wrong brand, or when someone asks whether displaying a brand's logo is allowed.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Don't lecture an owner who has already reasoned about this.

    line 30
  2. Distinctive:

    Run it from a page on your own domain, not from a file:// page and not from curl.

    line 81
  3. Distinctive:

    Verify by counting what rendered, in a real browser:

    line 122
  4. Very specific:

    When a favicon service can't resolve a domain it returns a default globe, with HTTP 200.

    line 146
  5. Distinctive:

    Issuers vs. products: Use the company's domain, not a product marketing page.

    line 163

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Brand logos, without rehosting anything

A row labelled "United MileagePlus" reads faster with United's logo beside it. Most people solve this badly: right-click a logo off a press kit, commit the PNG, watch it rot.

This is the pattern that works, and the traps that cost real deploys.

The legal shape (know it before you refuse)

Using a company's logo to identify that company — not as decoration, not as your own branding, not implying endorsement — is nominative use. Every comparison site, review site and directory on the internet relies on it. A logo beside a row about that company's product is the textbook case: the mark is doing referential work.

Two genuinely different acts, worth separating:

  • Displaying a mark to identify a brand. Routine. This is what you're doing.
  • Downloading a brand's asset file and redistributing copies from your server. Different — you're making and hosting copies of a file you don't hold rights to.

This skill avoids the second entirely. Nothing is scraped, nothing is committed. Each mark is fetched at runtime, by the visitor's browser, from the brand's own domain. You are not in the distribution path.

Don't lecture an owner who has already reasoned about this. Engage the actual distinction, or build.

The chain — first hit wins

1. logos/<id>.svg        ← file the owner dropped in (self-hosted override)
2. logos/<id>.png        ← same, PNG
3. DIRECT[<id>]          ← the brand's OWN /favicon.ico, from the brand's domain
4. favicon proxy         ← https://www.google.com/s2/favicons?sz=64&domain=<domain>
5. brand-coloured monogram (UA, MB, DL…)

Steps 1–2 let any single logo be upgraded later without touching code. Step 4 renders for most brands. Step 5 means a missing logo is never a broken image.

⚠️ Step 3 (DIRECT) usually FAILS. Read this before you use it — I shipped it and it broke.

The idea is sound: the proxy is a cache, and it is sometimes wrong or stale. In a real 39-brand run it returned a generic globe for one airline, and a 16px mush for a retailer whose own domain served a scalable SVG. So: go straight to https://<brand>/favicon.ico. Better quality, and the mark comes from the company itself.

It does not work, and the reason is invisible to your tests.

curl  https://www.evaair.com/favicon.ico   → 200, valid icon    ✅
curl  https://www.costco.com/favicon.ico   → 200, valid SVG     ✅

<img src="https://www.evaair.com/favicon.ico">  from another origin  → FAILS TO PAINT ❌
<img src="https://www.costco.com/favicon.ico">  from another origin  → FAILS TO PAINT ❌

Hotlink protection. Those servers see a cross-origin Referer and refuse. curl sends no Referer, so it sails straight through. Every check you'd naturally run says the URL is fine. The browser says otherwise, silently, and your logo drops to a monogram.

I verified those URLs with curl, shipped them, and two brands fell back to monograms in production. A 200 is not a painted logo.

If you use a DIRECT entry, verify it with an actual image load, from a different origin:

const test = src => new Promise(res => {
  const i = new Image();
  i.onload  = () => res({ src, ok: true, w: i.naturalWidth });
  i.onerror = () => res({ src, ok: false });          // ← hotlink block lands here
  i.src = src;
  setTimeout(() => res({ src, ok: false, note: 'timeout' }), 8000);
});

Run it from a page on your own domain, not from a file:// page and not from curl. Most brands will fail. Treat DIRECT as a rare exception you have proven, not a default.

const LOGO_DOMAIN = { united: "united.com", marriott: "marriott.com", /* … */ };
// DIRECT: only for brands you have PROVEN paint via a real <img> load from your own origin.
// Most brands hotlink-block and will silently fail here (see the warning above).
// The two examples I originally shipped BOTH failed in production. Left empty on purpose.
const DIRECT = {};
const favicon = d => `https://www.google.com/s2/favicons?sz=64&domain=${d}`;

function logoFallback(img, id, mono) {
  const step = img.dataset.tried || "svg";
  if (step === "svg") { img.dataset.tried = "png"; img.src = `logos/${id}.png`; return; }
  if (step === "png" && DIRECT[id]) { img.dataset.tried = "direct"; img.src = DIRECT[id]; return; }
  if ((step === "png" || step === "direct") && LOGO_DOMAIN[id]) {
    img.dataset.tried = "favicon"; img.src = favicon(LOGO_DOMAIN[id]); return;
  }
  img.remove();                       // nothing resolved → monogram
  const m = document.getElementById(mono);
  if (m) m.hidden = false;
}

THE TRAP: loading="lazy" silently kills the whole chain

Never put loading="lazy" on these images.

An offscreen lazy image never attempts to load. No load attempt → onerror never fires → the entire fallback chain does nothing → you ship a grid of invisible empty tiles that look fine in code review and broken to a user.

Someone will add it later while "optimising images." Leave a comment saying why it must not be there.

The second trap: verifying with a status code

  1. curl without -L. The favicon endpoint 301s. Without -L you'll conclude every domain failed and start fixing things that work.
  2. Checking for HTTP 200. A 200 says the server answered. It doesn't say a logo was painted, and it doesn't say it's the right brand.

Verify by counting what rendered, in a real browser:

[...document.querySelectorAll('.plogo')].map(m => {
  const img = m.querySelector('img'), mono = m.querySelector('span[id^=mono]');
  if (img && img.complete && img.naturalWidth > 0) return "LOGO " + img.naturalWidth + "px";
  if (mono && !mono.hidden) return "monogram";
  return "PENDING/BROKEN";     // ← any of these = broken chain (usually lazy-loading)
});

Then look at them. Build a contact sheet — every mark at real display size — screenshot it, and actually look. That is the only way to catch a logo that resolved to something that isn't the brand.

Monograms must be unique — check for collisions

The fallback monogram is usually an airline/loyalty code. Two brands can legitimately share one. We shipped BR for both EVA Air (its IATA code) and Bilt Rewards — two unrelated brands rendering the same two letters, which is worse than no logo at all.

Assert uniqueness across the whole map before you ship. It is three lines and it catches a real bug.

Detect the generic globe — the silent wrong answer

When a favicon service can't resolve a domain it returns a default globe, with HTTP 200. It looks like a logo. It isn't. Fingerprint it and compare — scripts/check-favicons.sh does this.

Styling that matters

  • White tile behind the mark: background:#fff; padding:2px; border-radius:5px. Most brand marks are drawn for light backgrounds; on a dark UI they vanish or look filthy.
  • 22px inline, ~34px header-sized. object-fit:contain.
  • Restructure the row/pill/table if it fights the logo. Don't wedge.

Sourcing domains: the part that needs care

A wrong domain gives a globe, or someone else's logo. Watch for:

  • Programs whose domain differs from the parent brand. Flying Blue is flyingblue.com, not airfrance.com. Lufthansa's program is miles-and-more.com.
  • Subdomains that serve worse icons than the parent. all.accor.com → 16px; accor.com → 64px.
  • Issuers vs. products. Use the company's domain, not a product marketing page.

Report quality honestly

Some brands publish bad favicons. The owner needs to know which to replace. From the real run:

What you'll findWhat to say
Only a 16×16 favicon exists (two major airlines, one retailer)Soft at 22px. Not fixable by you. Needs a hand-sourced SVG.
The favicon is a wordmark, not a symbol (a hotel chain's "FOR THE STAY")Fine at 34px, mush at 22px.
Resolves to something ambiguous (an orange ring that doesn't read as the brand)Flag it. Don't ship it silently.
Proxy returns a globeUse DIRECT, or accept the monogram.

Do not quietly ship something fuzzy and call it done. The pattern's virtue is that it degrades honestly, and any single logo can be fixed by dropping a file in logos/ — no code change.

Reference

  • scripts/check-favicons.sh — fetches every domain (with -L), reports real format and pixel size, and flags any that came back as the generic globe.
  • scripts/logo-contact-sheet.html — throwaway page rendering every mark at 22px and 44px on a dark background. Serve it, screenshot it, look at it.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.