Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Produce a creative brief grounded in live reference sites and deliberately distinct from existing demos, for any new brand build. Composes with brand-archetype-system and creative-direction. Use this skill at the start of any brand or microsite build when the result needs to land in a specific aesthetic position and not drift into a default house style. Triggers on creative brief, design brief, build brief, brand direction, where should this brand sit, what should this site look like, reference
.claude/skills/rampstackco-creative-brief-selector/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 121% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 88% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 71% | 0% |
A starting-point selector for brand and microsite builds. Given a business spec (name, vertical, shape, one-line vibe, optional list of shipped demos), produces a creative brief grounded in three things:
brand-archetype-system, picked or composed to be deliberately distinct from the shipped demos.The brief output is concrete tokens, not abstract families, including explicit section-shape choices (hero shape and footer shape) as first-class outputs alongside palette and voice. It is meant to be loaded by downstream build skills as the single creative artifact for the build.
creative-direction).logo-design, brand-identity).brand-voice).brand-ideation first, then return to this skill).This skill works upstream of build-time skills and parallel to the aesthetic methodology layer:
brand-archetype-system (upstream input). The archetype catalog this skill picks from. The skill names the archetype in the rendered brief; it does not redefine the archetype's defaults.creative-direction (parallel input). The four-axis vocabulary the archetypes reference. If the consumer needs the full axis brief, run creative-direction first; this skill's brief consumes its outputs.brand-identity (downstream). Once the brief is approved, brand-identity turns it into a finished identity system.landing-page-copy, content-and-copy, art-direction (downstream). Every downstream skill that produces aesthetic output references this skill's brief.Direct verbatim copying of an archetype's default palette into a new brand is the failure mode this skill exists to prevent. The brief shifts the archetype's defaults toward the business spec's specifics.
brand-archetype-system that fit the vertical and shape.references/reference-bank/, load the file(s) matching the chosen archetype-and-vertical. Augment with one to two discovered live references if the bank is sparse. Add discovered references back to the bank in the build PR.references/02-brief-template.md. Run output-side divergence. Output the brief plus the references list plus the divergence-check result.Full step-by-step in references/01-process.md. The divergence schema and its three overlap rules are in references/03-divergence-check.md.
A curated bank that grows with each build. Each file under references/reference-bank/ covers one archetype-and-vertical combination and holds three to six live reference URLs, each with a one-line why and optional palette or type observations. The bank ships with three seed combinations covering western-boot maker, heritage barbershop, and balloon-ride experience.
When a build draws from a sparse combination, the build is expected to commit the live references it discovered back to the bank as part of its build PR. The bank gets denser with use.
Each shipped demo carries a signature with seven fields:
slugarchetype (the canonical archetype name from brand-archetype-system)dominant_hue_family (the recognizable colour family, e.g. leather-bone-saddle, dawn-navy-coral, dark-linen-amber)voice_register (e.g. story-forward third-person, atmospheric second-person, fitment-first technical)primary_structural_pattern (e.g. shoppable-grid-product-forward, arc-timeline-hero, fitment-selector-then-rails)hero_shape (e.g. dual-column-image-and-text, wide-photograph-with-band-below, full-bleed-image-with-overlay; canonical vocabulary in references/05-section-shapes-vocabulary.md)footer_shape (e.g. single-line-strip, multi-column-sitemap, type-only-no-links; same vocabulary file)Overlap rules (full set in references/03-divergence-check.md):
A worked example file of seven signatures lives at references/04-shipped-demos-signatures-example.md for illustration; the live signatures file for any consuming project lives in that project, not in this skill.
Archetypes are NAMED for aesthetic families, NOT for brands, following the convention of brand-archetype-system. Live reference sites are cited as exemplars using nominative attribution language: "exemplified by URL]", "characteristic of", "in the register of". This is descriptive and nominative fair use territory, durable across brand redesigns. A brand that pivots its identity does not invalidate the archetype it once exemplified.
This skill's output depends on data, measurements, or tool results it cannot generate on its own. When a required input, tool, or data source is unavailable or unverifiable, the sanctioned output is the deliverable with the gap stated: what was needed, what was actually obtained or verified, and which parts of the output are affected. Fabricating, estimating, or interpolating a required number to complete the deliverable is never sanctioned. A stated gap is a complete answer.
references/00-overview.md - The case for the skill, the hybrid references model, the two-direction divergence, the trademark posture.references/01-process.md - The five-step process, each step concrete enough to execute.references/02-brief-template.md - The fillable brief template with section-by-section guidance.references/03-divergence-check.md - The signature schema and the overlap rules.references/04-shipped-demos-signatures-example.md - Worked example signatures file with seven demos.references/05-section-shapes-vocabulary.md - Canonical open vocabulary of hero and footer section shapes plus archetype affinities.The curated reference bank lives in a reference-bank/ subdirectory under references/. The bank ships with a README plus three seed archetype-and-vertical files: premium-dtc-maker-western-boots.md, heritage-local-service-barbershop.md, and hospitality-experience-balloon-ride.md. The bank's purpose and extension procedure are documented in its README; the seed files each carry three to four positive live references and one to two negative references for the chosen position. Load the bank file matching the build's archetype-and-vertical at step 3 of the process.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 38,302 | 22,898 | -40% | 1 | 1 | 0% | 6,124 | 5,966 | -3% | 0 | 0 | — |
case-02 | fail→pass | 23,966 | 24,435 | +2% | 1 | 1 | 0% | 4,066 | 6,343 | +56% | 0 | 0 | — |
case-03 | fail→fail | 21,396 | 22,504 | +5% | 1 | 1 | 0% | 3,638 | 5,641 | +55% | 0 | 0 | — |
case-04 | pass→pass | 8,995 | 8,583 | -5% | 1 | 1 | 0% | 1,467 | 3,613 | +146% | 0 | 0 | — |
case-05 | fail→pass | 11,113 | 6,518 | -41% | 1 | 1 | 0% | 1,861 | 3,378 | +82% | 0 | 0 | — |
case-06 | pass→pass | 6,291 | 7,671 | +22% | 1 | 1 | 0% | 1,096 | 3,314 | +202% | 0 | 0 | — |
case-07 | pass→pass | 14,170 | 9,892 | -30% | 1 | 1 | 0% | 2,126 | 3,805 | +79% | 0 | 0 | — |
case-08 | pass→pass | 12,872 | 6,926 | -46% | 1 | 1 | 0% | 2,026 | 3,317 | +64% | 0 | 0 | — |
case-09 | fail→pass | 7,944 | 4,322 | -46% | 1 | 1 | 0% | 1,285 | 2,844 | +121% | 0 | 0 | — |
case-10 | fail→pass | 18,081 | 15,780 | -13% | 1 | 1 | 0% | 2,554 | 4,798 | +88% | 0 | 0 | — |
case-11 | pass→pass | 16,403 | 8,035 | -51% | 1 | 1 | 0% | 2,306 | 3,430 | +49% | 0 | 0 | — |
case-12 | fail→pass | 14,251 | 9,433 | -34% | 1 | 1 | 0% | 2,100 | 3,585 | +71% | 0 | 0 | — |
case-13 | pass→pass | 13,946 | 15,529 | +11% | 1 | 1 | 0% | 2,028 | 4,468 | +120% | 0 | 0 | — |
case-14 | pass→pass | 12,148 | 8,071 | -34% | 1 | 1 | 0% | 1,780 | 3,436 | +93% | 0 | 0 | — |
case-15 | pass→pass | 7,525 | 4,300 | -43% | 1 | 1 | 0% | 1,212 | 2,847 | +135% | 0 | 0 | — |
case-16 | pass→pass | 11,047 | 8,178 | -26% | 1 | 1 | 0% | 1,694 | 3,461 | +104% | 0 | 0 | — |
case-17 | pass→pass | 16,923 | 22,981 | +36% | 1 | 1 | 0% | 2,388 | 5,433 | +128% | 0 | 0 | — |
case-18 | fail→pass | 14,988 | 11,925 | -20% | 1 | 1 | 0% | 2,135 | 4,145 | +94% | 0 | 0 | — |
case-19 | fail→pass | 18,125 | 4,313 | -76% | 1 | 1 | 0% | 985 | 2,860 | +190% | 0 | 0 | — |
case-20 | fail→pass | 24,257 | 8,559 | -65% | 1 | 1 | 0% | 3,745 | 3,418 | -9% | 0 | 0 | — |
case-21 | fail→pass | 31,881 | 9,239 | -71% | 1 | 1 | 0% | 6,178 | 3,483 | -44% | 0 | 0 | — |
case-22 | fail→fail | 20,727 | 14,978 | -28% | 1 | 1 | 0% | 3,253 | 4,391 | +35% | 0 | 0 | — |
case-23 | pass→pass | 19,694 | 5,602 | -72% | 1 | 1 | 0% | 3,079 | 3,038 | -1% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +39 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.