Skip to content

gooseworks-ai/goose-skills / drafts/marketing/learn-my-brand

learn-my-brand

Verified

Learn my brand once and stop asking me every time. Use when someone says learn my brand, set up my brand, here's our brand book, remember how we talk, what can we claim, can we say this, approve our claims, sign off our voice, our price or product changed, the client approved the claims, pick up where we left off, or check our claims are still current. Reads what GooseWorks already found about the brand, asks only the gaps as drafts to confirm, audits every claim (including ones research marked approved) and gets the ones that matter signed off with proof, scope and review date, turns the voice into rules the owner confirms, checks logo, colours and fonts, and saves it all as it goes so every other job loads it instead of asking. Ends with an unknowns list and a read-back. Not for who our customers are and what they say (use research-audience) or writing positioning from scratch (not covered yet).

references/brain-map.md

View on GitHub
# Brain map: where each piece of brand knowledge lives today

Load at steps 1 and 7. This is the write-back contract: for each of the 23 knowledge sections, where the brand brain keeps it today, how this playbook reads it, and how it writes a confirmed answer back. Tool names are the GooseWorks MCP's; the live schema is the truth for parameters. Snapshot of what the server exposes as of 2026-10-16.

## How a write is made

Field names, limits and encodings come from the live schema; this page only says which home each answer goes to. The save protocol itself (order, record versions, read-back) is `atoms/brain-write-back`.

- **The owner said it, or approved the exact text** → `gooseworks:brand_update` with `knowledge_intent: user_correction` and `user_statement` holding only words the owner (or the approver, or the person in the chat for facts they state) typed: their statement, or an explicit approval that stands on its own. Never your draft, research text or a bare "ok". A relayed approval keeps its provenance: "<name> via <user>, email of <date>: '<their words>'".
- **Found or inferred, not confirmed** → leave it unwritten, or `knowledge_intent: agent_proposal` with a `rationale`. A proposal stays pending until the owner accepts it in the GooseWorks app; siblings don't see it.
- **Rules** → learnings (`patch.facts`), through `../../../../skills/marketing/save-learning/SKILL.md` (the only seam; it holds any interim handling of today's server): the server stores them as the user's. A lean `brand_read` returns every `must`/`dont`. One idea per fact, so each can be found, merged and retired (`atoms/brain-write-back` step 3).
- **Everything with no typed home** → the Brand record (below). Never a second copy of a typed field.

## The Brand record (the interim home)

`brand/records/<YYYY-MM-DD>-v<N>.md` in the brand coworker scope (`coworker_agent_id` from the `brand_read` list; no coworker → default workspace with paths prefixed `brands/<brand-slug>/`), imported as the brand document `Brand record — <Brand> — <YYYY-MM-DD> v<N>` through `../../../../skills/marketing/save-to-brain/SKILL.md` (`versioned: supersede`). It holds what the brain has no typed field for: claims with variants and proof, the sign-off sheet, not-approved and retired claims, never-say, business facts, contradiction decisions, palette and fonts, unknowns. Format: `references/brand-record-template.md`.

- Each save is a new version built from the previous one read in full with `gooseworks:file_read` (search snippets would drop rows). save-to-brain deletes the older versions' documents after the new import is ready; deleted media drops out of search. The workspace files stay as history.
- The marker names the current version. When the record and the `Claims you may use —` facts disagree, the facts win; the record is proof detail.
- This is an interim home until the server has typed homes for claims, never-say rules, unknowns and business facts.

## How siblings read this

Through `../../../../skills/marketing/claims-gate/SKILL.md` (the only claims contract; AGENTS §8) and the kit voice plus `../../../../skills/marketing/voice-card/SKILL.md`. In short: the marker's status token decides (`signed-off` → the `Claims you may use —` facts; anything else → plain product facts, or the gate's scoped sign-off for the lines the job needs); every `dont` and `must` always applies; variants and proof come from the record the marker names, read in full. The real fonts and full palette are in that record too.

## The 23 sections

Status today: **typed** = a real field holds it; **rule** = held as learnings facts; **record** = only the brand record document holds it; **server** = the server researches and owns it, this playbook only reads and confirms.

| # | Section | Read from | Write a confirmed answer to | Status |
|---|---|---|---|---|
| 1 | The business and its goal (model, stage, the number marketing owns, deadline, budget committed vs movable, biggest blocker) | `brand_read` summary + onboarding (role, ad-spend band, goals) | company type → `patch.kit.brand_type`; `brand_update` `patch.onboarding` for role / ad-spend band / platforms / goals; everything else (goal number, deadline, budget, stage) → brand record | onboarding typed, rest record |
| 2 | What we sell (products, prices, policies, hero offer, standing discounts, loyalty/referral/subscription programmes, what it does not do) | `brand_read` products lean, then `product_ids` for full records | `patch.products` per product (offers with end dates, notes such as hero and what it does not do, `ad_eligibility` + `ad_eligibility_reason`). Offers and policies the brand states in copy (free shipping, returns, a discount) → one `Claims you may use` fact each with their end or review date; operating rules ("never discount below 20%") → `Policy:`; an email-lifecycle `Standing offer:` fact is retired or rewritten when the owner changes the offer. All also in the record | typed per product + rule |
| 3 | What sells and what we can afford (bestsellers, margins, AOV, repeat rate, LTV, the most we can pay per customer) | nothing | brand record | record |
| 4 | Where people find and buy us (site, marketplaces, app stores, stockists, locations, countries, languages, shipping) | summary, products' links, `sources` | brand record | record |
| 5 | Who we sell to and who we don't | kit `audience` (prose) | kit `audience` via `brand_update` `patch.kit` (siblings read the kit) | typed |
| 6 | What customers say | `knowledge_search` `source_types: ["evidence"]` | not written here; research-audience writes the voice-of-customer bank (`research/audience/voc-bank.md`). Review imports (`knowledge_sync`, paid) are offered, never run without a yes | server |
| 7 | Competitors and alternatives | `competitor_read` tracked + `view: suggestions` | `gooseworks:competitor_save` add for each rival or alternative the owner names (free; starts background research). Pre-revenue: ask what buyers would use instead, including nothing. Never remove one here (it deletes research; the owner confirms on the card) | typed |
| 8 | Search terms | nothing | not asked here; improve-search-visibility asks | unknown by design |
| 9 | Why people pick us (category, promise, differentiators, message pillars, objections, tagline, boilerplate) | kit `description`, `valueProps`, `tagline`, `messagingAngles`, `existingAds` | kit `description` and `value_props` (only approved claims); `patch.knowledge.positioning` (owner's own one-liner) and `.tagline` | typed |
| 10 | Our story and beliefs | kit `description`, About page via research | brand record (founder story, origin, beliefs, topics on and off limits) | record |
| 11 | How we sound | kit `voice` (prose), `tone`, `aesthetic`; owner-made learnings; voice cards in `brand/voices/` | kit `voice` holds the confirmed guide (personality, rules with pairs, terminology, flex, and the full list under "Banned:"); the record's Voice section repeats the banned list; one `Banned words —` fact (owner's words first); voice status → the marker; per-account or per-person style → `../../../../skills/marketing/voice-card/SKILL.md` | typed + one rule |
| 12 | Our look and files | kit logos (+ `logoConfidence`), palette, `fonts`, `typography`, `referenceImages`, `productsNeedingImages`; `gooseworks:media_list` scope `brand` | primary and accent colour via `patch.kit`; logo files via `gooseworks:media_upload` (kind `logo`); the full palette and the real fonts → brand record (no writable field yet) | typed in part (server-owned kit) |
| 13 | Claims and proof | kit `approvedClaims`, `proofPoints`, `creativeAxes`; `knowledge_search` `["claim"]` (all candidates) | **No tool writes or changes a claim.** One `Claims you may use —` fact per approved claim + the record (variants, proof, the not-approved list); the marker says everything else is not approved (see `atoms/brain-write-back`, `../../../../skills/marketing/claims-gate/SKILL.md` §5) | rule + record (no claim-writing tool yet) |
| 14 | Rules we can't break (regulated category, platform rules, banned claim types, disclaimers, consent, the owner's never list, disclosure, trademarks) | learnings | one `Never say —` fact per line, tagged law, platform or owner; the full table with reasons in the record | rule + record |
| 15 | How we work together (who approves claims, voice and spend, what the agent may do alone, spend caps, who to call in a crisis) | nothing readable (the Operating Plan exists in the app but is not readable or writable here) | record: the claims and voice approver; spend approver, crisis contact and spend caps go to the unknowns ledger as "asked when you run ads / when something goes wrong" unless the owner volunteers them | record |
| 16 | Creators, partners, press | nothing | not asked here; work-with-creators asks | unknown by design |
| 17 | Accounts and tools | `brand_read` accounts; Meta connection via `meta_read` (not needed here) | brand record lists which tools exist and who owns them; never ask for passwords | record |
| 18 | How we count results | nothing | not asked here; analyze-performance agrees definitions once | unknown by design |
| 19 | This week and what's coming (focus, live promotions, launches, next big date, what not to promote) | nothing | brand record (next big date, current focus, anything not to promote and until when) | record |
| 20 | Our plan | `campaign_read` (plan-campaign writes it) | not written here | other playbook |
| 21 | What we still need to know | brand record unknowns ledger | brand record | record |
| 22 | Everything we've run | server records GooseWorks work and Meta history | not written here | server |
| 23 | What's working | learnings (`insight` facts) | not written here; analyze-performance writes learnings | other playbook |

## Rules every written fact follows

- **Provenance.** Every line in the brand record carries its source (told by owner / found on site / found in research / inferred), the date, and who confirmed it.
- **Public or not.** Mark anything internal, embargoed (with the date) or under NDA. Default for numbers like revenue, margin and budget: internal.
- **Scope.** Say what a fact applies to: the whole brand, one product, one audience, one market.
- **Unknown is a value.** A gap is written as unknown with the job it blocks and the assumption in use, never filled with a guess.
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall