From supervibe
Use WHEN design-facing agents need retrieval-backed style, UX, brand, deck, chart, collateral, or stack UI evidence TO ground decisions in project memory, code facts, approved design-system tokens, and the internal design intelligence data pack. Internal support only; no user-facing slash command.
npx claudepluginhub vtrka/supervibe --plugin supervibeThis skill is limited to using the following tools:
Internal lookup and synthesis support for Supervibe design work. This skill does not own brand direction, UX specs, prototypes, presentation decks, accessibility review, or stack implementation. It supplies cited evidence so those agents make better decisions.
data/app-interface.csvdata/charts.csvdata/collateral/cip-deliverables.csvdata/collateral/cip-industries.csvdata/collateral/cip-mockup-contexts.csvdata/collateral/cip-styles.csvdata/collateral/icon-styles.csvdata/collateral/logo-colors.csvdata/collateral/logo-industries.csvdata/collateral/logo-styles.csvdata/colors.csvdata/google-fonts.csvdata/icons.csvdata/landing.csvdata/manifest.jsondata/products.csvdata/react-performance.csvdata/slides/slide-backgrounds.csvdata/slides/slide-charts.csvdata/slides/slide-color-logic.csvMandates invoking relevant skills via tools before any response in coding sessions. Covers access, priorities, and adaptations for Claude Code, Copilot CLI, Gemini CLI.
Share bugs, ideas, or general feedback.
Internal lookup and synthesis support for Supervibe design work. This skill does not own brand direction, UX specs, prototypes, presentation decks, accessibility review, or stack implementation. It supplies cited evidence so those agents make better decisions.
Use this skill only as internal evidence support for design-facing agents and commands that need retrieval-backed product, style, UX, chart, deck, collateral, or stack UI guidance. It should enrich a design decision with citations, not replace the owning design, review, or handoff skill.
Follow docs/references/skill-expert-operating-standard.md: start from source
of truth, preserve retrieval evidence, apply scope safety, use real producers
with runtime receipts for durable delegated outputs, verify before completion
claims, and keep confidence below gate when evidence is partial.
Before lookup, read the active design workflow state, approved design-system
manifest if present, project memory, Code RAG hits for existing UI/tokens, and
docs/references/design-expert-knowledge.md. For regulated-trust briefs, gather
domain evidence before accepting palette, typography, copy-risk, or trust
defaults. Apply the Reference Quality Ladder before allowing any external
reference to influence a recommendation. For new products, rebrands, missing
systems, material visual shifts, or weak category references, also read
docs/references/creative-reference-taxonomy.md and the relevant local packs in
skills/design-intelligence/references/creative/ before recommending creative
benchmarks.
Approved design system exists and matches the target
-> Reuse tokens, components, decisions, and prior approval evidence.
Design system is candidate, needs_revision, missing, or mismatched
-> Return a gap and route to the owning design-system workflow before durable output.
Local memory, code, or design-intelligence rows conflict
-> Apply precedence: approved design system, project memory, codebase patterns, accessibility law, external lookup.
Brief is regulated-trust or safety-sensitive
-> Require domain evidence before palette, typography, trust-copy, or data-display recommendations.
Lookup has no relevant local evidence
-> Return fallback reason and recommend the owning skill gather external/current evidence if needed.
Use through existing routes only:
/supervibe-design for design-system, prototype, presentation, collateral, and stack handoff work./supervibe-audit for UI polish, accessibility, token drift, brand asset drift, and deck quality review./supervibe-strengthen for agent tuning and repeated design failure patterns./supervibe routing when the trigger router selects a design intent.Never add a new slash command, package script, or standalone CLI wrapper for this lookup.
Required preflight order:
design, brand, ux, a11y, tokens, prototype, slides, or rejected.designContextPreflight() or searchDesignIntelligence() for the relevant domains.docs/references/creative-reference-taxonomy.md, then select 1-3 local packs
from skills/design-intelligence/references/creative/. Use packs to produce
borrow/avoid moves and differentiation pressure, not brand imitation.referenceRole, qualityTier, capturedAt, borrow, avoid, and notAuthority=true unless it is an approved project brand source.Before design-facing outputs, read docs/references/design-expert-knowledge.md
and apply Design Pass Triage from its Eight-Pass Expert Routine where the
owning command or agent is doing substantial design work. The owning workflow
classifies each pass as required | reuse | delegated | skipped | N/A. This
skill supplies the local evidence lookup pass; it does not replace preference
intake, reference scan, IA/user-flow, visual-system, responsive/platform,
quality, or feedback/approval passes.
When an approved design system already exists, treat prior preference and visual-system choices as reusable evidence unless the user asked for a rebrand, new audience posture, or material direction change. When a candidate or needs_revision design system exists, recommend resuming section approval instead of treating it as reusable prototype evidence. If lookup reveals a missing token, component, asset, or interaction, recommend a narrow design-system extension instead of a full restart. Do not force all eight passes for every prototype when local evidence proves reuse is sufficient.
Design-facing outputs must include a compact Design Intelligence Evidence section when lookup influenced a recommendation:
Design Intelligence Evidence:
query: "<user/design query>"
memory:
- path: ".supervibe/memory/decisions/example.md"
relevance: "accepted or rejected design decision"
lookup:
- id: "style:12:example"
domain: "style"
score: 3.42
recommendation: "retrieved row summary"
conflict: "none | project memory wins | design system wins"
references:
- source: "https://example.com/path"
referenceRole: "creative benchmark | interaction benchmark | category convention | direct competitor | platform standard | implementation library | anti-pattern | do-not-use-as-style"
qualityTier: "tier-1 | tier-2 | tier-3"
capturedAt: "YYYY-MM-DD | local-pack"
borrow: "specific trait or pattern to reuse"
avoid: "specific trait, risk, or mimicry to avoid"
notAuthority: true
creativePacks:
path: "fast path | medium path | full creative path"
selected:
- "skills/design-intelligence/references/creative/creative-data-products.md"
borrow: "specific pack moves used"
avoid: "specific pack risks rejected"
acceptedStatus: "candidate | accepted | rejected | learned"
fallbackReason: "no matching row | stack data unavailable | external lookup skipped"
Use the ladder from docs/references/design-expert-knowledge.md for all
non-project references. A source can be a creative benchmark, interaction
benchmark, category convention, direct competitor, platform standard,
implementation library, anti-pattern, or do-not-use-as-style. Famous product
names must be decomposed into borrow/avoid traits before use; they are not
style authority.
Tier-1 references need a current capture date or official version, direct fit,
and explicit borrow/avoid notes. Tier-2 references are advisory. Tier-3
references are historical context or anti-pattern material only.
Local creative reference packs are tier-2 evidence. For a full creative path,
select multiple packs with genuinely different moves so the owning agent can
generate distinct candidate directions instead of small variations of the same
style.
Apply this hierarchy without exceptions:
Generic retrieved guidance is advisory. It cannot override approved tokens, prior explicit rejection, accessibility requirements, or codebase facts.
product, style, color, typography, ux, landing, app-interface: primary design and UX guidance.charts: chart choice, fallback, and accessibility guidance.icons: icon library/import guidance, adapted to Supervibe's local icon policy.google-fonts: font pairing and font availability evidence.react-performance: UI performance rules for React surfaces.ui-reasoning: decision framing, critique, and recommendation quality evidence.stack:*: implementation guidance for React, Next.js, Vue, Svelte, Angular, Flutter, SwiftUI, Shadcn, Tailwind, Three.js, and related stacks.slides:*: presentation strategy, layout, copy, chart, typography, color, and background guidance.collateral:*: logo, icon, CIP, brand asset, and mockup context guidance.creative packs: local tier-2 pack cards in
skills/design-intelligence/references/creative/ selected through
docs/references/creative-reference-taxonomy.md.Domain aliases stack, slides, and collateral expand to the corresponding
prefixed local domain families during lookup.
When a design-facing agent handles a product, prototype, landing, deck, or
collateral request, it must consider every relevant local dataset family before
claiming that evidence is unavailable: product, style, color, typography, UX,
landing, app-interface, charts, icons, google-fonts, react-performance,
ui-reasoning, stack, slides, collateral, and creative packs. If a family is not
relevant, mark it N/A with rationale; do not omit it silently.
For advanced visual work, pair local evidence with a Prototype Capability Plan. Library families agents may consider include native CSS/WAAPI, Motion, GSAP, Lottie/lottie-web, Rive, Three.js, PixiJS, D3, Observable Plot, ECharts, MapLibre GL, Theatre.js, Rough.js, Matter.js, Monaco, CodeMirror, and stack-specific chart wrappers such as Visx or Recharts for handoff. The local pack helps select fit and risk; it does not grant permission to import a dependency without the prototype capability gate.
Use the local knowledge pack first. Do not instruct design agents to fetch a remote repository or remote skill for baseline expertise.
For design-facing lookups, map retrieved evidence to this coverage matrix before making a recommendation:
Use a product-fit style matrix to match product category, trust/risk, density, platform, interaction mode, and data intensity before suggesting visual direction. Use stack-aware UI guidance for framework, mobile, or component-library handoff; retrieved style rows are advisory and cannot override approved tokens, accessibility requirements, or local code patterns.
Write memory only after a review or user signal marks the result accepted, rejected, or learned. Do not write every candidate suggestion.
Accepted decisions need artifact links or evidence ids. Rejected alternatives need a rejection rationale so future agents do not resurrect the same weak option.
Returns:
supervibe:brandbook - owns approved design-system creation.supervibe:prototype - consumes evidence for UI prototypes.supervibe:presentation-deck - consumes deck and slide evidence.supervibe:ui-review-and-polish - verifies design evidence in rendered artifacts.