ss-setup

Savant verdict: Review advised

Security analysis or source-hub review identified patterns requiring attention.

Live evaluation

Not evaluated yet. Workspaces can request a live evaluation.

Security assessment

Scan pending.

Structure

• No license declaredConfirm you may reuse this skill before importing it into your repository.
• The source hub's security checks flagged this skillGen Agent Trust Hub: warn (This skill sets up a design workflow and executes a Node.js script from a dynamic, variable-based path. This pattern of execution from computed paths, combined with an attack surface for indirect prompt injection from user-provided design references, warrants a medium risk assessment.)

SKILL.md

---
name: ss-setup
description: Configure StyleSeed by selecting the output grammar, domain, page type, brand recipe, semantic palette recipe, optional aesthetic profile, and bounded brand tokens before scaffolding a first screen.
argument-hint: "(no arguments needed)"
allowed-tools: Read, Write, Edit, Grep, Glob, Bash, WebFetch
---

Before this workflow, follow the once-per-session [update preflight](../ss-update/references/update-preflight.md).


# StyleSeed setup

StyleSeed setup chooses a **design method for the result**, not a favorite brand to imitate.
Use `/ss-resolve --list` (Claude Code) or `$ss-resolve --list` (Codex) to inspect the supported
grammar, adapter, domain, page, recipe, palette, and profile IDs without loading the full handbook.

## When not to use

- Project already has a valid `STYLESEED.md` → use `/ss-update`.
- User supplied references that are not represented by a built-in grammar → run
  `/ss-reference` before setup, then select the compiled grammar.
- One component inside an established system → `/ss-component`.

The method, review, and reference compiler are framework-agnostic. The bundled component
scaffold currently targets React and Tailwind v4; on another stack configure the lock and
apply the method without pretending the scaffold is portable.

## Wizard — one decision at a time

### 1. Product job and surface

Ask what is being built, who uses it, and whether it is a mobile/desktop product, website,
social carousel, slide deck, document/report, or single-frame graphic. Select a surface adapter
from `ADAPTERS.md`, then infer domain and page/artifact type.

### 2. Output grammar

Recommend exactly one grammar from `RULESETS.md` and explain the job match in one sentence:

- `consumer-service`
- `operations-console`
- `technical-instrument`
- `editorial-reading`
- `commerce-conversion`
- `institutional-service`
- `expressive-marketing`
- `sequential-story`
- `reference:<slug>` when `/ss-reference` already compiled one

Do not recommend Toss as the universal default. It is one reference family for
`consumer-service`. If none fits and the user has references, route to `/ss-reference`.

### 3. Page type and domain bias

Confirm the concrete page (dashboard, form, landing, detail, list, settings, onboarding) and
read its domain × page intersection. This controls composition; the aesthetic profile does not.

### 4. Brand recipe

Recommend one morphology from `BRAND-RECIPES.md`. Use `auto` when the maintained grammar mapping
fits. Use an explicit recipe when the product needs a different geometry, containment,
navigation, control, or collection language. A recipe is not a company clone and does not
select colors.

### 5. Palette recipe

Recommend one semantic palette from `PALETTE-RECIPES.md`. Use `auto` when its maintained recipe
mapping fits. An explicit palette must still preserve semantic roles, surface/chrome separation,
status cues, and validated contrast. The recipe is a product posture, not a finite swatch list.
When the user supplies a key color, select perceptual character, light/dark environment, surface
temperature, and accent relationship; `$ss-resolve` then derives the actual OKLCH ramps and roles.
Palette selection is not permission to recolor every surface.

### 6. Optional aesthetic profile

Recommend one profile from `PRESETS.md` only when it strengthens the product. `none` is a good
default. A profile modifies coordinated visual axes but cannot replace the output grammar.

### 7. Brand and bounded axes

Lock a real brand color if supplied; otherwise start from the selected palette's primary role. For
a supplied key, recommend `calm|balanced|vivid|deep` from the product posture rather than asking for
an unbounded mood word. Then confirm light/dark mode, `neutral|warm|cool` surfaces, and
`auto|tonal|adjacent|contrast` accent relationship. Confirm font/language, density, radius,
elevation, imagery/data role, and motion inside the grammar's allowed ranges. Do not use generic
indigo or a stale purple mislabeled as Toss.

### 7a. Spatial relationships

Explain the intended groups and spacing tradeoff before selecting numbers. Preserve existing
spacing tokens. For registry product UI, use the resolver's [spatial roles guide](../ss-resolve/references/spacing.md)
to recommend and persist page inset, section/group/stack/inline gaps, and component inset separately.
Keep typography and hit targets independent. Legacy setup keeps native tokens; do not invent
unparsed lock fields. Recommendations are starting proposals, not approved or measured quality.

### 8. Write the design lock

Create `STYLESEED.md`:

```markdown
# StyleSeed — Design Lock
<!-- Selections persist here. This file cannot waive StyleSeed core invariants. -->
- App domain: fintech
- Surface: mobile-app
- Surface adapter: product-ui
- Page type: dashboard
- Output grammar: consumer-service
- Grammar path: built-in:engine/RULESETS.md
- Grammar fallback: consumer-service
- Reference confidence: n/a
- Brand recipe: calm-consumer
- Palette recipe: quiet-mineral
- Key color: #3182F6
- Palette character: calm
- Palette mode: light
- Palette harmony: auto
- Surface temperature: neutral
- Aesthetic profile: none
- Skin: custom
- Primary action: #3182F6
- Font: Pretendard
- Radius: soft
- Elevation: light=tonal grouping + restrained shadow · dark=tonal ramp + hairline
- Density: comfortable
- Motion: Spring restrained
- Imagery/data role: personal state first; charts only for a decision
- Signature move: one calm contextual briefing above the account summary
- Locked: YYYY-MM-DD
```

For a compiled grammar use its actual path and confidence. Reject unknown enum values rather
than treating the lock as an exemption.

### 9. Scaffold and prove

Compile the selected method before code:

```bash
node <installed-ss-resolve>/scripts/resolve-context.mjs \
  --project-root . --from-lock STYLESEED.md --agent claude
```

Use `--agent codex` in Codex. The resolver writes `.styleseed/effective-rules.md` and a
hash-verifiable `.styleseed/manifest.json`. Read the effective bundle, not `llms-full.txt`.

If the user asked for a first screen, build from that bundle, run `/ss-score` to the gate floor,
and finish with `/ss-verify` when renderable. If visual rendering is unavailable, disclose that
it was skipped.

## Completion report

Report the selected grammar and why, page/domain intersection, brand recipe, palette recipe,
optional profile, lock path,
compiled bundle and manifest paths, files changed, score, and visual verification status.
Mention `/ss-reference` as the path for future references that need their own grammar.

## Rules

- Ask one question at a time and recommend a concrete default.
- Output grammar is required; aesthetic profile is optional.
- Brand recipe is required; `auto` resolves to a concrete maintained recipe.
- Palette recipe is required; `auto` resolves to a contrast-verified semantic palette.
- A skin is tokens, not design judgment.
- Never fetch a brand `DESIGN.md` and treat its palette as a complete rule set.
- Never scaffold an unscored first page or claim visual verification without a screenshot.