| name | atomic-brand |
| description | Audits web projects against atomic design discipline and brand-token coherence. Detects token violations, atomic hierarchy drift, naming by appearance, scattered components, brand drift, and scale violations. Three verdicts: system-healthy (ship), drifting (refactor against plan), broken (stop; build missing tokens and patterns first). Includes failover chain for brand detection: explicit tokens → parse image → scrape live URL. Activates on "atomic audit", "brand audit", "design system audit", "audit my components", "is this on-brand", "find duplicate components", "consolidate components", or "/atomic-brand". Auto-suggests when the user pastes a URL with a visible `tokens.css` or `tailwind.config` nearby.
|
atomic-brand
Purpose
Sit at the design-system layer. Block scattered components, hardcoded
values, brand drift, and naming-by-appearance before they ship. Emit a
scored audit, a refactor plan (when drifting), or a build-out plan (when
broken).
Web medium first. Native (Swift, Compose), TUI, print, and email are
candidates to become sibling skills when migration triggers fire — see
TESTS.md.
Modes
Audit mode (default)
Trigger: "atomic audit", "brand audit", "/atomic-brand"
Behavior:
- Detect brand-token source via the failover chain
- Walk the component library (CSS, HTML, React/Vue/Svelte files)
- Apply the 8-dimension rubric per
references/audit-rubric.md
- Emit the Atomic Brand Audit with findings
- Issue verdict; emit downstream artifact (refactor plan or build-out plan)
Refactor mode
Trigger: "find duplicate components", "consolidate components"
Behavior:
- Run audit
- Focus the refactor plan on the highest-leverage consolidations and renames
- Emit
refactor-plan.md ordered by leverage
Build-out mode
Trigger: automatic when verdict is broken
Behavior:
- Identify which tokens and patterns are missing
- Emit
build-out-plan.md listing required tokens, atoms, molecules, organisms
- Sequence the build by T-shirt size (XS → XL) where applicable
The failover chain (brand detection)
Three sources of truth, tried in order:
| Step | Source | Confidence |
|---|
| 1 | Explicit token file (tokens.css, tailwind.config.{js,ts}, tokens.json, theme.ts) | high |
| 2 | Brand asset image (logo, screenshot, style guide PDF) | medium |
| 3 | Scrape live URL — extract computed styles from rendered pages | low |
If step 1 finds a token file, use it. Skip 2 and 3.
If step 1 fails, try step 2.
If step 2 fails, try step 3.
If all three fail, verdict = broken; build-out plan starts with "no
brand tokens defined."
See references/failover-chain.md.
The 8 audit dimensions
| # | Dimension | What it scores |
|---|
| 1 | Token compliance | % of color/spacing/type values using tokens vs. hardcoded |
| 2 | Atomic discipline | Honest hierarchy: atoms compose into molecules compose into organisms |
| 3 | Naming integrity | Components named by role, not appearance |
| 4 | Brand coherence | Visual identity consistent across the surface |
| 5 | Scale discipline | Values come from defined scales (spacing, type, color) |
| 6 | Composition health | No cross-tree reaches; clean dependency graph |
| 7 | Duplication | Minimum number of components per job |
| 8 | Variant clarity | Variants are declared and named, not implicit |
Each scored 1-5. Verdicts: system-healthy (all ≥4), drifting (any 2-3),
broken (any 1 OR ≥3 dims at 2).
See references/audit-rubric.md.
Failure-mode taxonomy
Every finding gets one tag from this set:
atomic_violation · token_violation · naming_violation ·
brand_drift · inconsistency · composition_violation ·
scale_violation · duplication_violation
See references/failure-modes.md.
Verdict gates
| Verdict | Allowed downstream |
|---|
system-healthy | Ship |
drifting | Apply refactor plan, re-audit |
broken | Stop; apply build-out plan; re-audit when minimum tokens exist |
Operating flow
1. Detect brand-token source (failover chain)
2. Walk component library; classify each file
3. Score each of 8 dimensions
4. Tag findings by failure mode
5. Compute verdict
6. Emit canonical Atomic Brand Audit (assets/audit-template.md)
7. Based on verdict:
- system-healthy: done
- drifting: emit refactor-plan.md
- broken: emit build-out-plan.md
References
references/atomic-design.md — Brad Frost's atomic design distillation
references/brand-tokens.md — token taxonomy by category
references/composition-rules.md — dependency-graph rules; what imports what
references/audit-rubric.md — 8-dimension scored rubric
references/failover-chain.md — brand detection ordering
references/failure-modes.md — tagged violation taxonomy
Prompts
prompts/audit-url.md — orchestrator; runs the full flow
prompts/audit-component-library.md — codebase walk and classify
prompts/audit-token-compliance.md — token vs. hardcoded scoring
prompts/audit-brand-coherence.md — visual identity audit
prompts/parse-image-for-brand.md — failover step 2
prompts/scrape-for-brand.md — failover step 3
prompts/refactor-plan.md — drifting verdict downstream
prompts/build-out-plan.md — broken verdict downstream
Triggers
| Phrase | Mode |
|---|
atomic audit, brand audit, /atomic-brand | Audit |
design system audit, audit my components | Audit |
is this on-brand, brand drift | Brand coherence focus |
find duplicate components, consolidate components | Refactor |
what tokens am I missing, build out my design system | Build-out |
Interfaces
| Layer | Convention |
|---|
| Color tokens | Reads from --color-* CSS custom properties via project token files |
| Type tokens | Reads from --font-* CSS custom properties; defers chart-specific or type-scale-specific audits to dedicated skills if present |
| Output formats | Markdown audit + JSON conforming to assets/audit-report-schema.json + optional refactor-plan.md or build-out-plan.md |
| Sizing | All emitted tasks use T-shirt sizes; consumers can re-size against their own rubric |
Scope (v0.1.0)
Web only. CSS, HTML, React, Vue, Svelte component libraries. Native and
other mediums spin out as sibling skills when migration triggers fire.
See TESTS.md for the trigger table.
Self-application
This skill must pass the canonical skill-structure audit at ≥4 on every
dimension (spec-compliant + marketplace-ready).