How the portfolio's Brand engine is structured and why it's built this way — the charters, the rules, the mechanics, and the tools each seat uses. Written for Robert, who asked how our system works so he can replicate it his own way.
DRAFT · for Stacy's review before it goes to RobertOne brand terminal used to serve the whole portfolio across every medium — brand books, logos, websites, social, email, video, all of it. Demand outgrew a single seat. So Brand became a hub plus specialists: one hub that owns the brand system and is the front door for every new brand, and a set of spokes that each own one craft and go deep on it.
Why it works: building great websites is a different craft from designing a logo, which is different from cutting video. Splitting them lets each spoke get genuinely good at one thing, while the hub guarantees they all still look like the same family. It also scales — you add a spoke when a craft gets busy enough to need its own seat, not before.
One hub. Nine spokes today — five brand-craft, four web-build. The list isn't closed; you add a spoke when a craft earns one.
| Layer | Terminal | Owns, in a phrase |
|---|---|---|
| HUB | brand-creative | The brand system + front door for new brands |
| Craft spokes | brand-copy | Website & landing-page words |
| brand-video | Motion, ad cuts, thumbnails, exports | |
| brand-social | Social + thumbnail creative (phone-legible) | |
| brand-email | Email template & campaign design | |
| brand-dashboard | Dashboard design system (skins the data) | |
| Web-build spokes | web-traveladvisors | Advisor sites (agency core + advisor clients) |
| web-retreathosts | Retreat-host sites & retreat pages | |
| web-retreatvenues | Venue booking-first sites | |
| web-general | Everything else on the web |
The list isn't fixed — in our system a spoke gets added when a craft gets busy enough to earn its own seat, not before.
brand-creative; "brand-core" is the role, not a rename.Each spoke: what it owns, what it explicitly does not, and the flow it sits in. Copy these as the seed for each terminal's CLAUDE.md.
All four exist for one reason — Rule 0: centralize the build (§5). Production websites get built only by these specialists, to the Brand quality bar. They split by audience so each gets deep at one kind of site.
Brand makes it · Ads runs it · Analytics measures it. The hub + spokes are the "makes it" half, now specialized. Brand and Analytics are horizontal services any project can call; Ads runs the paid side.
Standing up any terminal — hub or spoke — is the same five ingredients. Get these and the three shared capabilities come for free.
~/Claude/<name>/).CLAUDE.md in that folder — the terminal's charter + house rules. This is the whole brain of the seat. (Skeletons in the appendix.)~/Claude/handoffs/handoff-<name>.md — a "here's what you are, freshly stood up" note so the first session has context.CLAUDE.md — so the CoS knows the seat exists and can route to it.handoffs/handoff-<name>.md (session continuity).A shared bridge file (~/Claude/bridge/terminal-bridge.md) is the peer-to-peer message log. Every message is FROM-<SEAT>, optionally TO-<SEAT>, tagged [fyi] / [ask] / [task] / [answer] / [decide]. Seats auto-act on most tags and only surface [decide] to the human. Work items get logged on the relevant project's board in the shared ops hub (ELMO), tagged by the spoke that did the work (e.g. [brand]).
A charter says what a seat is for; this is the toolkit each seat draws on to do the work. Shared as a reference — for context on how our seats are equipped, not a checklist for Murph. (/skill = a slash-command skill; MCP = a connected tool server; CLI = a command-line tool.)
| Seat | Skills (/slash) | Tools (MCP / CLI) |
|---|---|---|
| brand-creative (hub) |
brand-guidelines · frontend-design · artifact-design · dataviz · skill-creator (to build the hub's own skills) | Canva (brand kits, brand templates) · Google Drive/Workspace (store the brand system) · Firecrawl (reference research) |
| brand-copy | content-research-writer · customer-research · marketing-research · product-marketing-context · website-benchmark (the AEO/SEO surface) · page-cro | Ubersuggest (keyword/SERP/AEO data) · Firecrawl (SERP + competitor scrape) · Google Docs (deliver the copy doc) |
| brand-video | ffmpeg-usage · pickframes (best thumbnail frame) · video-downloader · youtube-transcript | ffmpeg (CLI — the whole cutting/rendering engine) · Canva (motion templates) · Playwright (frame grabs) · TikTok (short-form reference) |
| brand-social | ui-ux-pro-max · brand-guidelines · pickframes | Canva/Vista (the production surface) · TikTok (format/trend reference) · Playwright · Firecrawl |
| brand-email | email-sequence · internal-comms · brand-guidelines | Kit (templates/broadcasts) · the ESP's own API (Brevo/Kit) · Google Workspace |
| brand-dashboard | dataviz · ui-ux-pro-max · frontend-design | PostHog (product/traffic data) · Ubersuggest (SEO data) · Playwright (view the live dashboard) |
| web-* (all four builds) |
frontend-design · ui-ux-pro-max · framer + framer-code-components · deploy + deploy-preflight · run · verify · website-benchmark (QA the built site) | GitHub · Cloudflare (wrangler CLI — the house Pages stack) · Netlify · Wix · Framer agent · Playwright (drive + screenshot + verify) · MarkUp.io (client feedback — pull the exact pin location from the API before interpreting a comment) · Firecrawl (scrape existing sites for rebuilds) |
That's the whole system. How Murph stands it up — which seats, which infrastructure, on what timeline — is Robert's to shape however fits Murph.
CLAUDE.md skeleton# Brand / Creative — Hub (brand-core) [MURPH]
**Role:** Owns the brand SYSTEM and is the front door for new brands.
Brand books, voice, palettes, logos, design tokens, cross-brand consistency.
When a NEW brand needs an identity, it is born HERE. Hands specs DOWN to the
spokes; is the source of truth they pull UP from.
## Charter
- OWNS: brand books, voice, palette, logo, design tokens, the design system,
cross-brand consistency, the client design-diversity system.
- DOES NOT OWN: the builds (web-* spokes), finished channel assets (craft
spokes), positioning/offer/pricing (business unit/client), product eng.
## Rules
1. Specialists pull the system; they never invent it. Unclear spec = a spoke
asks the hub, doesn't guess.
2. New brand -> hub first (no book/tokens yet = born here).
3. Home Base routes; the CEO never picks a brand terminal.
4. Make-vs-deliver seams: social MAKES/socials POST · email DESIGNS/ESP SENDS ·
dashboard DESIGNS/Analytics DATA · copy WRITES/web BUILDS · web BUILDS/unit DIRECTS.
5. Honor house rules: bigger-by-default fonts, external-comms gate (drafts free,
SEND is the human's), never paste secrets in chat.
## Baseline (free): /handoff · /gr · /look · bridge identity = BRAND · ELMO [brand] tag
CLAUDE.md + seed handoff skeleton# <Spoke Name> (<slug>) — a spoke under the Brand hub [MURPH]
## What this terminal is
<one-sentence definition of the craft>
- YOU OWN: <the one craft>
- NOT YOURS: identity (hub) · other spokes' mediums · the DIRECTION
(business unit/client) · <anything adjacent>
- PULL THE SYSTEM from the hub (brand-core) before making anything.
Missing/unclear spec -> FROM-<SLUG> TO-BRAND [ask]. Never invent it.
## The flow you sit in
<unit/client DIRECTS -> you MAKE -> (web-* BUILDS if you're copy) -> hub owns VOICE>
## First-run
1. Read the hub contract + the relevant brand book(s).
2. /gr to sync the bridge; introduce yourself: FROM-<SLUG> [fyi] "live".
3. Await routed work from Home Base (or a [task] from a unit / another spoke).
## Baseline (free): /handoff · /gr · /look · bridge identity = <SLUG> · ELMO tag
The seed handoff is just the same intro saved to handoffs/handoff-<slug>.md with a "freshly stood up — no carry-forward yet" note, so the first session opens with context.