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/voice-craft.md

View on GitHub
# Voice craft

Load at step 5. The brain usually holds the server's voice read as a prose paragraph plus tone adjectives. A paragraph cannot reject a sentence, so siblings write generic copy from it. This step turns it into rules a writer can apply and a stranger can check, then gets the owner's yes. Do not re-distill the voice from scratch: start from the brain's read and the brand's own lines, then sharpen.

## What the confirmed voice contains

1. **Personality: is / is not.** 3–4 pairs, each with a tension: "Confident, not cocky", "Expert, not clinical", "Warm, not gushy". Each side gets one real sentence.
2. **3–5 voice rules, each falsifiable.** A rule, then "we say" (a real brand sentence) and "we don't say" (a *plausible* off-brand version someone would actually write, not a straw man). Shape, with illustrative lines; always use the brand's own:
   - *Lead with the feel, not the chemistry.* We say: "Light enough for morning, gone in ten seconds." We don't say: "Our proprietary complex leverages advanced actives."
   - *Specific over superlative.* We say: "Fragrance-free. 30 ml lasts about two months." We don't say: "The ultimate game-changing formula."
   - *Talk like someone who's been there.* We say: "We tried everything too. Most of it stung." We don't say: "We understand your skincare journey."
   Every "we say" line is itself copy: it must pass the approved-claims list and the never-say list before it becomes a rule (no efficacy promise, nothing that addresses the reader's own health, body or finances — Meta bans that).
3. **Banned words and patterns.** The owner's own list, plus this standing list of words and patterns that make copy read as machine-written: elevate, unlock, unleash, seamless(ly), game-changer, revolutionise, delve, embark, journey (as metaphor), empower, supercharge, level up, curated, robust, holistic, transformative, "in today's fast-paced world", "look no further", "say goodbye to", "whether you're X or Y", "it's not just X, it's Y", "Imagine…" openers, rhetorical-question openers, reflexive groups of three, em-dash chains, emoji bullet lists. The owner may keep any word on purpose; record that.
4. **Terminology.** Brand and product name styling (capitalisation, ™/®), what customers are called, words the brand owns, words it avoids, competitor naming policy.
5. **Flex by channel.** One line each for the channels the brand uses: ad hook (punchier, one claim), email (warmer; from the founder or the brand?), social (how loose; slang yes or no), customer support (plain, apologise once, no hype, never promise outcomes), B2B sales (concrete numbers, nothing cute), app-store or marketplace listing (plain, scannable).
6. **Flex by speaker.** Founder voice ("I": opinionated, personal; swears or not) vs brand voice ("we"). run-organic-social and make-content depend on knowing the difference.
7. **Mechanics.** Locale and spelling (US/UK/other), reading level (consumer: plain, roughly grade 6–8), headline case, numerals, Oxford comma, exclamation marks (cap per asset), emoji (which, where, never in support).
8. **Taste notes.** Lines the owner loves or hates from anywhere, each with their reason in their own words.

## Where the samples come from

Pull 6–10 real lines across channels before asking anything. Trust order:
1. Founder-written (posts, emails, interviews, the About page if they wrote it).
2. Best-performing (own ads by results, top emails by clicks; `existingAds` hooks are a start).
3. Recent (last 12 months).
4. Agency or freelancer copy.
5. Site copy. Often agency- or AI-written: check it against the banned list before trusting it.

Discard any line that would fit the top two competitors equally well; a rule that fits them too is not voice. If the brand has almost no copy (pre-launch), ask the founder to paste 5–10 of their own posts or messages and use those (plus their words in this chat) as the sample; say the voice is provisional.

**Owner-made rules beat the research read.** If a learning the owner wrote (e.g. "don't write clipped short sentences in scripts") clashes with the research voice ("sentences are short"), show the clash and let the owner settle it; never keep both.

## Getting the owner's yes efficiently

1. **Triad test.** Take one real brand line and rewrite it three ways along one dial (plainer / warmer / punchier). Ask: "Which is us?" Then a second dial (formal ↔ casual, or playful ↔ serious). Two or three rounds settle most of the guide.
2. **Keep or kill.** Show five short lines written in the draft voice: an ad hook, an email subject line, a social caption, a support reply, and (B2B) a sales line. The owner marks keep or kill and says why for any kill. Each "why" becomes a rule, in their words.
3. **Show the guide** (rules with pairs, banned words, terminology, flex) as one screen and ask for corrections. Their yes, verbatim, is the sign-off.

Never ask "describe your brand in three adjectives", "what's your brand personality?", or a rating scale. Show, then ask them to choose.

## Failure modes to check before showing the guide

- Adjective soup with no rules ("bold, friendly, innovative").
- A rule that cannot reject any sentence.
- A voice copied from the category leader, or distilled from AI-written site copy.
- No channel or speaker flex, so support replies sound like ads.
- Straw-man "don't" examples.

**Falsifiability test:** give someone one rule and two sentences; they should be able to pick the on-brand one.

## How it is written back

- The confirmed guide (personality, the rules with their pairs, terminology and name styling, mechanics, channel and speaker flex) → the kit's `voice` via `brand_update` `patch.kit`, as a `user_correction` with the owner's approval words. Siblings read the kit voice. Keep it within the field's limit by writing tight pairs, not prose.
- Banned words and patterns → the full list (the owner's own words plus the standing list, minus words they keep on purpose) under "Banned:" in the kit `voice` and in the Brand record's Voice section; plus one `Banned words — …` fact (`kind: "dont"`) holding the full list, the owner's own words first. The rest of the voice guide lives in the kit `voice`, its home, not in learnings.
- The voice status (`confirmed` or `provisional`) goes in the claims marker; provisional voice is kept in the Brand record only.
- A founder's or account's own style → `../../../../skills/marketing/voice-card/SKILL.md` (`brand/voices/<handle>.md`). The brand's rules beat a card's style.
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: 196kUpdated Claude CodeCodexVerifiedInstall