Your files, and where everything lives

There are two file trees involved: yours (your brand data, which the toolkit writes into) and the toolkit’s (the reference material it ships with, which never changes). This page covers yours. What’s in the toolkit covers the other one.

The one rule

Everything durable hangs off your brand root. In a single-brand workspace that is brand/. In a multi-property workspace each client gets properties/{name}/.

That means moving a client between machines, or handing one off to someone else, is copying one directory. It carries the Brain, every ledger, the customer-voice corpus, and every dated deliverable with it.

brand/                        (or properties/examplebrand/)
  brand-guide.md              your profile. Every skill reads this first
  brain/                      the 29-module knowledge hub, indexed
  ledgers/                    one line per run, per skill, so history is queryable
  voc/                        customer-voice corpus
    raw/                      your original exports. APPEND-ONLY, never rewritten
  deliverables/{skill}/       dated run outputs

The 11 brand data files

These live at your brand root and are the source every skill reads before producing anything.

File	What it powers
`brand-guide.md`	Visual identity, positioning, offers, and your conversion shape. Read by everything
`voice-dna.md`	Your writing voice. Used by every copy skill
`psychographic-map.md`	Creative angles across psychological layers
`hooks-by-awareness.md`	Hooks mapped to awareness levels
`buyer-personas.md`	Your 5 micro-personas
`objection-map.md`	Objections and their reframes
`value-props.md`	Tiered value propositions
`competitors.md`	Competitive landscape
`case-studies.md`	Your verified results. You provide these
`testimonials.md`	Your approved quotes. You provide these
`offer-deep-dive.md`	Offer structure, pricing, guarantees

brand-context populates most of these from your URL. The last three need you, because nobody can research your real numbers and your approved quotes into existence.

Why `voc/raw/` is never rewritten

Your raw customer exports are the only thing in this tree the toolkit did not author, which is exactly what makes them evidence.

A customer-voice bank that verifies itself against an index the same run wrote proves nothing; it will render a beautifully clean provenance table over completely invented quotes. So the raw exports are append-only, and the provenance check reconciles against them, not against the index. A run that regenerates a raw export has destroyed its own evidence, and its verdict becomes UNVERIFIABLE rather than clean.

The Brain

If you have run client-brain or business-brain-builder, brain/ holds the deep version: 29 numbered modules with an index that routes skills to the right ones rather than loading all of them.

Two modules earn special mention:

Module 20, compliance. Your banned claims. Record them once and every skill enforces them.
Module 27, access. Which ad account, pixel, analytics property, store and CRM belong to this brand, by identifier. This is what lets an audit confirm it is reading your account.

The ledgers

Skills that render a verdict append one line per run to ledgers/. Each line carries the date, the verdict, which rows passed or failed, which rows were NO-DATA and what was missing, and how long the verdict stays valid.

That last part is why “was this clean last month, or did nobody check?” is an answerable question rather than a guess.

What this means for you day to day

Back up your brand root. It is the whole of a client’s durable state.
Fill case-studies.md and testimonials.md early. Several skills stall politely without them, and none of them will invent proof.
Never hand-edit voc/raw/. Add files, never rewrite them.
Toolkit updates never touch your brand data. Different tree entirely.

What’s in the toolkit · Where numbers come from · Onboard your brand