brief

---
name: brief
description: Use when the user wants to create or edit a naming brief, provides a filled template, or wants custom naming rulesets. Prefills existing context, collects missing preferences through a short choice-first survey, confirms the brief, and saves it. For an end-to-end naming search, use find-domain.
---

# Brief

The brief is the subject matter of the brand. Every later step (`generate`, `research`, `language-check`, `report`) reads it. Use the fixed schema as a storage contract, not a mandatory question-by-question interview. Read [references/quick-survey.md](references/quick-survey.md) for the default intake and host-native choice controls. Save only after confirmation.

## Tools

- `ruleset_list`, `ruleset_get`, `ruleset_save`: rulesets to pick from, and custom rulesets.
- `project_create`: save a new project from a confirmed brief.
- `project_list`, `project_get`, `project_update_brief`: find an existing project and save a new version of its brief.

## Rules
- Preserve creative intent beyond product keywords. Save the emotional promise in `positioning.emotionalPromise`, directions and learned preferences in `nameRules.creativeGuidance`, and examples with reasons in `nameRules.references` (`name`, `preference`: `like`, `dislike`, or `benchmark`, `reason`). Benchmarks are not templates to imitate; rejected examples are not global bans on styles.
- Save explicit domain constraints in `nameRules.domainPolicy`: `tlds` is an exclusive list (exact .com means `["com"]`); `allowAffixes` and `includeRegistered` default to false in the skills. Set either true only when the user wants fallback spellings or registered/aftermarket domains. These choices override ruleset defaults, country-first TLDs, and extra TLDs. Do not ask again when already specified.
- Do not select a coinage-heavy ruleset just because the user wants a short exact .com. Choose rulesets compatible with the creative breadth; price tier does not determine naming style.
- Deliver results directly in chat by default. Files and visual review surfaces require a user request; project storage is internal continuity.

- Prefer clickable choices, in batches of up to three questions, using the host's available question interface. Fall back to numbered choices in chat. Use free text only when choices would force a guess or the user wants a custom answer. Ask only for information not already supplied.
- Never invent answers to required questions. Take them from what the user gave you, or ask.
- Do not save before the user confirms the summary.
- No emoji and no urgency copy.

## Step 1: Pick the input path

- The user gave a file path (usually a filled copy of `templates/brief.md` from this skill's folder): read it and go to Step 3.
- The user pasted a description of the product: go to Step 3.
- Otherwise run the choice-first intake in `references/quick-survey.md`. Step 2 is the field reference and optional guided interview, not the default sequence. If the user explicitly prefers a file, offer `templates/brief.md`.
- The user wants to change an existing project's brief: call `project_list`, let them pick, call `project_get`, and use `project.brief.data` as the starting answers. Ask what to change, then go to Step 5.

## Step 2: Interview

Use these fields to check completeness after the short survey. Run them as a sequential interview only if the user requests guided, one-at-a-time questions. R = required, O = optional. Optional questions may be skipped.

### 1. Subject
1. R `subject.what`: What is it, in one sentence?
2. R `subject.category`: What category or industry is it in?
3. R `subject.problem`: What problem does it solve, and for whom?
4. O `subject.features`: Key features or offer?
5. O `subject.keywords`: Descriptive keywords? Offer 5-8 keywords from the earlier answers to accept or edit.

### 2. Audience and markets
6. R `audience.who`: Who buys or uses it?
7. R `audience.segment`: Options: B2B (`b2b`), B2C (`b2c`), both (`both`).
8. R `audience.countries`: Which countries? Store ISO 3166-1 alpha-2 codes in uppercase (`DE`, `US`, `GB`).
9. R `audience.languages`: Which languages? Store BCP 47 codes (`en`, `de`, `pt-BR`). Suggest the main languages of the chosen countries.

### 3. Positioning
10. O `positioning.competitors`: 3-5 competitors, by name or URL (store at most 5).
11. O `positioning.difference`: What makes it different? Also capture `positioning.emotionalPromise` when supplied: how should the customer feel?
12. R `positioning.priceTier`: Options: budget, mid, premium.

### 4. Personality
13. R `personality.tone`: Three words for the tone, exactly three. Offer two or three sets that fit the earlier answers.
14. O `personality.archetype`: Options: Sage, Hero, Creator, Explorer, Outlaw, Magician, Lover, Jester, Everyman, Caregiver, Ruler, Innocent. Store it in lowercase.
15. O `personality.avoid`: Words or feelings to avoid.

### 5. Name rules
16. R `nameRules.rulesets`: Pick 1-3 rulesets. See Step 4.
17. O `nameRules.overrides`: Project-only length, include/ban words, and extra TLDs. Also capture `creativeGuidance`, `references`, and `domainPolicy` from supplied preferences. Do not turn optional creative context into a mandatory questionnaire.
18. O `nameRules.handles`: Which social handles matter? Options: github, npm, pypi, reddit, x, instagram, tiktok, youtube, linkedin.

### 6. Future
19. O `future.scope`: Will the brand grow past this product? Options: broad name (`broad`), narrow name (`narrow`).

### 7. Mode
20. R `mode.type`: Options: startup (`startup`), personal brand (`personal`), product (`product`).
21. Personal brand only, all required: `mode.fullName` (full name), `mode.profession` (profession), `mode.nameForms` (accepted name forms, for example "Jane Doe", "J. Doe", "Jane Doe Design").

If the first message already suggests a personal brand, ask question 20 right after question 1 so the later questions fit.

## Step 3: Template or free text

- Template: each question in the file has a `<!-- field.path -->` marker and an `Answer:` line. Map every non-empty answer to its field. Convert country names to ISO codes and language names to BCP 47 codes. Split comma-separated lists.
- Free text: fill every field the text supports. Leave out fields the text does not support; do not guess them.
- Then ask only for missing required facts and high-value preferences using the choice-first survey. Do not repeat answered questions or require the user to fill every optional field. If the brief is already complete, skip the survey.

## Step 4: Pick rulesets

Call `ruleset_list`. Suggest 1-3 rulesets with one line each on why. Starting points:

| Brief signal | Suggest |
|---|---|
| `mode.type` is `personal` | `personal` |
| Developer audience: CLI, API, library, open-source tool | `dev-tool` |
| Local service in one place | `local-business` |
| User explicitly wants compact invented names | `premium` or `coined`, after inspecting their styles |
| `priceTier` budget, or the lowest renewal cost matters most | `budget` |
| Plain names that say what it does | `descriptive` |
| A distinctive brand, open to new words | `coined`, `evocative`, `compound` |

Custom rulesets are in the same list with `builtin: false`. Suggest them when they fit.

If the user wants a custom ruleset:
1. Copy: call `ruleset_get` on a built-in, change what the user asks (length, styles, TLD order, `.com` policy, affixes, weights), give it a new lowercase id that is not a built-in id, and save it with `ruleset_save`.
2. From scratch: ask for `name`, `description`, 3-5 `examples`, `styles` with weights (descriptive, compound, invented, metaphor, misspelling, affix, person-name, place-service), `length` min and max, `filters` (hyphens, numbers, banned letter pairs, banned words), `tlds` (`order`, `countryFirst`, `com` as must, nice, or none, and `affixes` such as `get-` or `-hq`), and `weights` from 0 to 1 for length, say, spell, meaning, uniqueness, and price. Save it with `ruleset_save`.
3. If `ruleset_save` returns an error, explain it in plain words, fix the field with the user, and save again.

## Step 5: Confirm and save

1. Show a concise creative summary in chat: product truth, audience, emotional promise, personality, examples and rejection reasons, rulesets, and domain policy. Show missing required answers; avoid listing every skipped optional field.
2. Offer **Start the search** or **Adjust the brief** through the available choice interface. These confirm or correct the summary; a preselected option is not approval. If the user already explicitly approved this exact brief, do not ask again. For a brief-only request use **Save brief** instead of **Start the search**.
3. On confirmation:
   - New project: call `project_create` with the brief and a short `title` (the working name, or the product in 2-4 words).
   - Existing project: call `project_update_brief`. Tell the user this saves brief version N and that existing names keep the version they were made with.
4. If the tool returns an error, explain it in plain words, ask for the missing or wrong field, and call the tool again.
5. Tell the user the project slug and the next step, `generate`. When `find-domain` invoked this skill, it continues on its own.

## Brief shape

`project_create` and `project_update_brief` take this shape. Optional fields may be left out.

```json
{
  "subject": { "what": "...", "category": "...", "problem": "...", "features": ["..."], "keywords": ["..."] },
  "audience": { "who": "...", "segment": "b2c", "countries": ["DE"], "languages": ["de", "en"] },
  "positioning": { "competitors": ["..."], "difference": "...", "priceTier": "mid" },
  "personality": { "tone": ["warm", "practical", "playful"], "archetype": "caregiver", "avoid": ["..."] },
  "nameRules": {
    "rulesets": ["coined", "compound"],
    "overrides": { "minLength": 4, "maxLength": 10, "include": ["..."], "ban": ["..."], "extraTlds": ["app"] },
    "handles": ["instagram"]
  },
  "future": { "scope": "broad" },
  "mode": { "type": "startup" }
}
```

For a personal brand, `mode` is `{ "type": "personal", "fullName": "...", "profession": "...", "nameForms": ["..."] }`.

• SKILL.md 10 KB
• references/quick-survey.md 7 KB
• templates/brief.md 3 KB