guia-matthieu/naming-6-gates
Use when creating brand names, product names, or company names in 2026. Use when naming needs to work across languages, voice assistants, and AI recommendation engines. Use when previous name candidates failed trademark or had conflicts.
npx skills add https://github.com/guia-matthieu/clawfu-skills --skill naming-6-gatesBrand naming in 2026 is architectural engineering, not creative brainstorming. A name must be:
Core principle: Optimize for AEO (Answer Engine Optimization) first, aesthetics second.
digraph naming_process {
rankdir=TB;
node [shape=box];
brief [label="1. BRIEF\nDefine constraints"];
generate [label="2. GENERATE\nAI-assisted creation"];
filter [label="3. LINGUISTIC FILTER\nMultilingual stress test"];
legal [label="4. LEGAL FILTER\nTrademark screening"];
aeo [label="5. AEO FILTER\nEntity strength test"];
story [label="6. STORY FILTER\nNarrative potential"];
brief -> generate -> filter -> legal -> aeo -> story;
filter -> generate [label="fails" style=dashed];
legal -> generate [label="conflicts" style=dashed];
aeo -> generate [label="weak entity" style=dashed];
story -> generate [label="no narrative" style=dashed];
}
Define hard constraints BEFORE generating:
| Constraint | Target | Why |
|------------|--------|-----|
| Length | 7-8 letters max | Mobile icons, cognitive load, tokenization |
| Syllables | 1-3 syllables | Voice invocation, memorability |
| Character set | A-Z only | Domain availability, global keyboards |
| Forbidden sequences | ght, ough, silent letters | Pronunciation barriers |
Output format:
brief:
max_letters: 8
max_syllables: 3
target_markets: [FR, ES, US, UK]
nice_classes: [9, 35, 42]
avoid_phonemes: [ʒ, θ, ð] # Hard for non-native speakers
competitor_names: [Notion, Figma, Linear] # Style reference
forbidden_associations: [medical, financial, religious]
Use structured prompting with 5 typologies:
Words from unrelated domains applied to new context.
Abstract concepts made concrete.
New words with phonetic engineering.
Blend of two meaningful words.
Meaningless but phonetically pleasing.
Generation Contract:
CONSTRAINTS:
- Max 8 letters
- Pronounceable in French, Spanish, English
- No existing trademark in Class 42
- Distinct from: [competitor list]
OUTPUT FORMAT per candidate:
- Name: [word]
- Typology: [1-5]
- Phonetic: /IPA transcription/
- Syllables: [count]
- Rationale: [1 sentence]
Test each candidate across target markets:
Run through TTS engines in 15+ languages:
Verify no negative meanings in:
| Hazard | Example | Impact |
|--------|---------|--------|
| Silent letters | "Knight" | Voice assistant failure |
| Consonant clusters | "Strengths" | Non-native difficulty |
| Ambiguous vowels | "Read" (reed/red) | Confusion |
| Cross-language homophones | "Gift" (poison in German) | Negative association |
Kill criterion: If TTS mispronounces in >2 target markets, reject.
| Registry | Tool | Check For |
|----------|------|-----------|
| France | INPI data.inpi.fr | Exact + phonetic similarity |
| EU | EUIPO eSearch / TMview | Same Nice classes |
| US | USPTO TESS | Registered + pending |
| International | WIPO Global Brand Database | Madrid Protocol |
EUIPO "Early TM Screening" detects:
Kill criterion: Any conflict in target Nice classes → reject.
A name must be recognizable as a distinct entity by AI systems.
Prompt to test:
"What is [NAME]?"
GOOD response: "I don't have information about [NAME]" (clean slate)
BAD response: "[NAME] is a common word meaning..." (collision)
BAD response: "[NAME] could refer to several things..." (ambiguity)
Use Semrush/Ahrefs:
Kill criterion: If name collides with high-volume generic term, reject.
A name must be "tell-able" - capable of carrying a brand story.
For each candidate, verify you can build:
| Pillar | Question | Example Answer Required |
|--------|----------|------------------------|
| Origin story | "Why this name?" | Must have a genuine, shareable reason |
| Belief story | What does the brand stand for? | Values must connect to name |
| Product truth | What does it actually do? | Name should hint at function |
| Transformation | What changes for the customer? | Before/after must be nameable |
| Culture | What kind of community forms? | Name should enable identity |
Both work. But you must choose consciously and build accordingly.
Say the name aloud 10 times. Ask:
Kill criterion: If you can't tell a compelling origin story, reject.
| Assistant | Invocation Rules | Rejection Risks |
|-----------|-----------------|-----------------|
| Alexa | 2+ words required (unless trademarked) | Single generic words |
| Siri | Must be phonetically distinct | Confusion with system commands |
| Google | Strong Knowledge Graph presence needed | Names too similar to famous brands |
| Bixby | Avoid wake-word conflicts | "Open", "cancel", common verbs |
Test procedure:
For the final candidate, produce:
## [NAME] - Naming Dossier
### Basic Info
- **Name:** [word]
- **Typology:** [1-5]
- **Phonetic:** /IPA/
- **Length:** X letters, Y syllables
### Linguistic Clearance
- FR pronunciation: [pass/fail + notes]
- ES pronunciation: [pass/fail + notes]
- EN pronunciation: [pass/fail + notes]
- Cultural flags: [none / list]
### Legal Clearance
- INPI (FR): [clear / conflict]
- EUIPO (EU): [clear / conflict]
- USPTO (US): [clear / conflict]
- Domains available: [list]
- Social handles: [list]
### AEO Score
- Entity uniqueness: [1-10]
- Keyword difficulty: [score]
- Tokenization: [single/split]
### Narrative Potential
- Origin story hook: [1 sentence]
- Brand values connection: [1 sentence]
- Community identity: [suggested demonym]
### Voice Readiness
- Alexa: [pass/fail]
- Siri: [pass/fail]
- Google: [pass/fail]
### Recommendation
[APPROVED / NEEDS WORK / REJECTED]
[Final reasoning]
| Mistake | Why It Fails | Fix |
|---------|-------------|-----|
| Starting with brainstorming | Generates unusable names | Start with constraints (Gate 1) |
| Skipping TTS test | Voice assistant failures | Always test pronunciation |
| Ignoring Nice class overlap | Trademark conflicts | Check specific classes, not just exact matches |
| Choosing descriptive names | Weak AEO, hard to trademark | Prefer distinctive/invented |
| No narrative test | Brand can't tell its story | Always verify storytelling potential |
The 7-8 Letter Rule: Shorter = better for mobile, voice, memory, tokenization.
Phonetic Gold Standard: CVCV pattern (consonant-vowel-consonant-vowel) works globally.
AEO Priority: If AI can't cite you as a distinct entity, you don't exist in 2026.
Empty Vessel Strategy: Meaningless names (Kodak) require more brand-building but face fewer conflicts.
Voice-First: If Alexa can't invoke it, 8.4 billion assistants can't recommend it.
| Claude handles | You provide |
|---------------|-------------|
| Generating candidates by typology | Brief constraints (markets, classes) |
| Running linguistic stress tests | Final pronunciation judgment |
| Checking trademark databases | Legal professional validation |
| Evaluating AEO/entity strength | Strategic brand direction |
| Producing naming dossier | Final selection and approval |
name: naming-6-gates
category: branding
version: 2.0
author: GUIA
source_expert: 6-Gate Protocol (AEO-first naming)
difficulty: advanced
mode: centaur
tags: [naming, branding, trademark, voice-assistant, aeo, entity]
created: 2026-02-03
updated: 2026-02-03
Take guia-matthieu/naming-6-gates from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.