Skip to content

Latest commit

 

History

History
145 lines (91 loc) · 9.54 KB

AGENTS.md

File metadata and controls

145 lines (91 loc) · 9.54 KB

Equal Experts Brand Guidelines - Agent Instructions

This repository contains brand guidelines and design system rules for creating and reviewing Equal Experts branded content.

Reference and skills

Standards live in the passive reference library at reference/ — brand, design language, content, accessibility, and per-medium reference (slides, web-product, web-marketing, imagery, charts-data). Reference states the WHAT; it contains no loading taxonomy and no verbs. Every old kuat-docs/rules/... path maps to its new home in reference/MIGRATION-MAP.md.

Migration note (Phase 2 done): The global LOADING.md task→files taxonomy is retired and loading is now per-skill — each skill names the reference/ slices it needs. Procedure from the old tree (role cards, checklists, "before you create/review") now lives in the activity skills; the _to-skills/ holding area and the kuat-docs/rules/ redirect tombstones have been removed. Legacy kuat-review / kuat-create are rewired onto reference/ and keep working until Phase 5.

Activity skills (preferred): pick by job — create-web-app / review-web-app, create-imagery, create-presentation / review-presentation. Each carries its own intake + loads the reference/ slices it needs.

Intent (legacy review | create): Determine intent first; use skill kuat-review or kuat-create; resolve rules per skills/shared/resolve-rules.md.

Three layers: skills/ — session procedure; reference/ — compliance standards (the WHAT); consumer repo or npm package — implementation overlay / component docs.

Three entry points: Org (this repo, reference/); library (kuat-mono git + overlay); app (@equal-experts/kuat-react bundled agent-docs). See consumption-architecture.md.

Upstream vs Local Ownership (Consumption Contract)

When this repository is consumed by implementation repositories (for example kuat-mono), use this ownership split:

  • Upstream (kuat-agent-docs) owns: brand, foundations, content style, structure/pattern guidance, and task-to-context loading taxonomy.
  • Local implementation repos own: component architecture, package APIs, testing strategy, Storybook conventions, and contributor workflow.

Cross-repo load order

  1. Load upstream rules from this repository first.
  2. Load local implementation overlay rules from the consumer repository second.

Conflict resolution

  • If guidance conflicts on design/structure/content intent, upstream wins.
  • If guidance conflicts on implementation/API/testing/build behavior, local implementation repo wins.
  • If implementation behavior is still unclear, trust runtime evidence in this order: tests, Storybook behavior, package exports, source code.

External skills and tool guidance

External guidance (including shadcn-related skills/docs) is complementary. It may inform implementation details but must not override this ownership and precedence contract.

Platform isolation

Important: Type-specific rules should NOT reference each other.

  • ✅ Slides can reference foundations
  • ✅ Web/product can reference foundations
  • ❌ Slides should NOT need web/product rules
  • ❌ Graphics should NOT need web examples

This ensures each platform gets only relevant guidance.


Quick Reference (When Full Docs Unavailable)

Brand Colors

  • EE Blue: oklch(0.645 0.163 237.5) / #1795d4 - Primary actions, brand
  • Transform Teal: oklch(0.645 0.120 185.0) - Secondary actions
  • Tech Blue: oklch(0.435 0.090 240.0) - Navigation, structure
  • Equal Ember: oklch(0.625 0.200 65.0) - Warnings, highlights

Typography

  • Sans: Lexend
  • Mono: JetBrains Mono
  • Serif: Lora

Spacing

4px base unit. Scale: 4, 8, 12, 16, 24, 32, 48px.

Border Radius

  • Static content: 0px
  • Interactive (buttons): 6px
  • Form inputs: 4px

Accessibility

WCAG AA minimum contrast (4.5:1 for text, 3:1 for large text/graphics).


Behavior Guidelines

When working on Equal Experts content:

  1. Determine intent — review vs create; if unclear, ask
  2. Check documentation first before making design decisions
  3. Follow existing patterns - do not invent new ones
  4. Use semantic tokens - bg-primary not #1795d4
  5. Ask if unclear - request clarification rather than guessing

Review

  • Use skill kuat-review; run ensure-rules.sh when possible
  • Intake before findings (artifacts, review depth, output format; product context for web at product_ux/full)
  • Agree output format with the user; use Open questions when context is thin — do not invent user stories or research
  • Cite rule files for every violation

Create

  • Use skill kuat-create and type-specific "Before you create" pre-flight
  • Use checklists before delivery (e.g. slides checklist)

Full Documentation

See kuat-docs/README.md for complete documentation index.

Consumption architecture: kuat-docs/setup/consumption-architecture.md · Ownership: ownership-matrix.md


Design-system migration (active)

We are migrating from the old rules/ + intent-skills architecture to reference guidelines + activity skills + an Enterprise plugin. Plans live in docs/migration/. Run a phase with /kuat-phase <n>.

Core principles (apply throughout)

  • Reference is passive (the WHAT). A file belongs in reference/ only if it states what is true about EE. Anything that says how to do a job — verbs, role cards, checklists, "before you create/review", loading tables — is procedure and lives in a skill, not in reference. (The "passive test".)
  • Media types own their patterns. reference/media-types/<medium>/ (slides, web-product, web-marketing, imagery, charts-data); each holds a patterns/ subfolder describing how we do specific things in that medium. "Patterns" is not a top-level concept and is not technical detail.
  • Skills use progressive disclosure. Each SKILL.md stays short and links to the reference slices it needs — never inlines reference content, never relies on a global loading taxonomy.
  • Don't break current users. Legacy skills/kuat-create and skills/kuat-review keep working until Phase 5.

Conventions

  • One branch + PR per phase: migration/phase-<n>-<slug>.
  • Keep a dated decision log in docs/migration/LOG.md; record deviations and checkpoint decisions.
  • End every phase by filling docs/migration/report-phase-<n>.md from the template and outputting it.
  • Preserve git history on moves (git mv). Plan mode before edits; stop and ask on ambiguous high-impact decisions.

Distribution model (context)

Pinned semver + beta/stable channels + pre-registration via managed settings. A plugin release is how fresh reference reaches consultants. Bump version in plugin.json only (never also in the marketplace entry). See docs/migration/phase-3-plugin-packaging.md.

Contributor skills (Phase 7 — DS team, repo-local)

The skills under .claude/skills/ are contributor skills — for the small DS team who extend the system, not consultants who apply it. They are repo-local only: project-scoped, auto-discovered when working in this repo, and never packaged into a plugin, the marketplace, or managed settings (the build refuses to package from .claude/skills/, and verify-plugins.mjs asserts none leak into a payload). Being in the repo is the access control.

  • Colours change at the SoT, never by hand. reference/design-language/colours.md is a generated artifact (npm run tokens:generate from colors.tokens.json); a drift check (npm run tokens:check) gates it. Use generate-tokens.
  • Reference changes go through the gate. Add/edit reference with author-reference; gate it with review-reference-change (npm run reference:check — passive test, link integrity, structure, token drift).
  • New skills in house style: author-skill. Studio asset pack: prep-slides-master, curate-slide-layouts, add-brand-asset.
  • Component authoring (add-kuat-component) + the downstream variables.css/registry generation live downstream in kuat-mono — see docs/migration/phase-7-kuat-mono-handoff.md.

Contributing

How to propose, build, review, and release changes to Kuat — the hybrid model, the four sizes, and a per-surface "Proposing a…" page for each contributor skill + gate — lives in CONTRIBUTING.md (front door) and contribute/overview.md (full model). The contribute/ docs are governance/how-to and sit outside the passive reference/ library, so they are never bundled into a consumer plugin.