create-brand

Brand Identity & Design System — Orchestrator

Coordinates 8 specialized agents to transform product context into a brand narrative + AI-readable design system. Capability metadata (routes, prerequisites, load map, artifact contract) lives in routing.yaml. Agent table + per-route dispatch graphs + pattern-catalog map: references/agent-manifest.md. Methodology and history: references/playbook.md.

Core question: Does every visual decision trace back to who we are?

Critical Gates — load first

• No colors/fonts before strategy. Visual-agent runs parallel with strategy-agent; orchestrator verifies coherence at merge.
• No Layer 2 before Layer 1 completes. Token-architect needs visual-agent output; component-token needs token-architect output.
• Critic's cross-element coherence is mandatory. Radius↔archetype, type↔personality, color↔emotion — the matrix no individual agent can see.
• Stale upstream (>30 days) → generic archetypes. Recommend re-running research-icp if artifact dates are old.
• BRAND.md is prose, DESIGN.md is specification. Never mix registers.

Output

File	Audience	Register	Route
docs/forsvn/canonical/marketing/BRAND.md	Founders, marketers, copywriters, designers	Prose — brand book	A + B
docs/forsvn/canonical/marketing/DESIGN.md	AI coding agents, frontend engineers, design-system consumers	Specification — tables, formulas, exact values	B only
docs/forsvn/canonical/marketing/CREATIVE-DIRECTION.md	Art directors, photographers, campaign + render briefs	Art direction — mood, light, framing, motion (references tokens, never redefines)	B (+ refresh-only)
docs/forsvn/canonical/marketing/FRAME.md	Video/motion producers, short-form + full-frame social briefs	Frame direction — safe-area, type-at-distance, on-screen pacing, bumper grammar (references tokens, never redefines)	B (+ refresh-only)
docs/forsvn/canonical/marketing/ASSETS.md	Designers, art directors, asset producers, PMs	Checklist — GFM checkboxes with spec ref + target path	B only

ASSETS.md is deterministically projected from BRAND.md + DESIGN.md + declared platforms (Step 8.5) — auto-scans brand/ each run; human-owned [~] (in-progress) and [!] (blocked) markers preserved across runs. CREATIVE-DIRECTION.md and FRAME.md are both orchestrator-written (no new agent) — CREATIVE-DIRECTION names what DESIGN.md tokens mean; FRAME names how those tokens behave in a moving frame (safe-area, type-at-distance, pacing). Both are additive, refreshable standalone. Per-section format + frontmatter schema: references/format-conventions.md; FRAME.md depth: references/frame-direction.md. Optional visual renderings via Paper MCP, Claude Design handoff, or a brand-kit board (Step 9).

Quality Gate — critic check-groups

Four check-groups; full canonical checklist (13 BRAND + 13 DESIGN + ASSETS + 4 cross-file + narrative-tension NT-Q1-Q4 + brand-kit BK-G1-G9) in agents/critic-agent.md:

• BRAND.md — origin/values/voice/tagline/lexicon depth, no copywriting scope creep, narrative-tension gates.
• DESIGN.md — AI-readable header, font/iconography tables, per-theme palettes + semantic tokens, WCAG AA, archetype-justified --radius, surface/shadow/motion systems.
• ASSETS.md (Route B) — one section per declared platform, spec-ref'd + substituted paths, no invented/duplicated rows, [~]/[!] preserved.
• Cross-file coherence — radius↔archetype, type↔personality, color↔emotion, imagery↔archetype, voice↔visual, ASSETS≡BRAND≡DESIGN platform sets; AI-slop check via references/ai-slop-detection.md (0-1 clean / 2-3 review / 4+ regenerate).

Reference bar: references/example-brand.md + references/example-design.md (copy-paste / blind-build / competitor-swap / implementation-gap tests).

Before Starting

Apply references/_shared/before-starting-check.md. Recommended context:

Artifact	Source	Required?
research/product-context.md	research-icp	Strongly recommended — drives strategy + audience grounding
research/icp-research.md	research-icp	Strongly recommended — audience archetype + voice register
docs/forsvn/canonical/marketing/BRAND.md (existing)	prior run	Optional — re-run overwrites in place + bumps version:
docs/forsvn/canonical/marketing/ASSETS.md (existing)	prior run	Optional — Step 8.5 preserves [~]/[!] markers across re-runs
docs/forsvn/experience/{product,audience,brand,business,technical}.md	any skill	Optional — persisted answers for the 7 Pre-Dispatch dimensions

Pre-Dispatch + Mode

Run canonical Pre-Dispatch (references/_shared/pre-dispatch-protocol.md). Needed dimensions: product (1-line), audience, competitive landscape (3-5 names), voice intuition (3 adjectives or reference brand), aesthetic intuition (3 visual references), target platforms (mandatory enumeration — drives ASSETS.md), positioning intent. Warm/Cold Start prompts + 13-platform catalog: references/procedures/pre-dispatch.md.

Mode (references/_shared/mode-resolver.md): auto-downgrade for ≤3 sentences AND no prior artifacts → Route A. --fast forces Route A regardless. --fast does NOT skip Cold Start (especially Q6 platform enumeration), Critical Gates 1-5, or hard-block conditions. Session execution profile (single-vs-multi): inherit per references/_shared/execution-policy.md. Step 9 visual renderings bind the design target tool — inherit tool_targets or ask once per references/_shared/tool-target.md; tool-agnostic (no rendering, 9c) stays the default.

Routing

Ask: "Full brand system or quick brand for MVP?"

• Route A — Quick Brand (MVP): Step 0 → Layer 1 parallel (strategy + visual color/typography-only, logo deferred) → critic (strategy↔visual coherence only) → re-dispatch on FAIL (max 2 cycles) → deliver + "Run full brand-system when ready to build the design system."
• Route B — Full Brand System: Step 0 → Layer 1 parallel (4 agents) → Merge → Layer 2 sequential (3 agents) → critic → Step 8.5 ASSETS projection → Step 8.6 orchestrator writes CREATIVE-DIRECTION.md (art-direction synthesis; critic checks token coherence) → Step 8.7 orchestrator writes FRAME.md (frame-direction synthesis from DESIGN.md + CREATIVE-DIRECTION.md; critic checks atoms-sacred / no-redefinition) → Step 9 (optional Visual Renderings) → Step 10 deliver. Max 2 rewrite cycles.
• Refresh-only: on a locked brand, (re)write CREATIVE-DIRECTION.md or FRAME.md standalone (read BRAND/DESIGN + CREATIVE-DIRECTION + any inspiration/ frames; bump that one file's version:; touch no other canonical file — regenerate the named artifact alone, never BRAND/DESIGN). Schema + section order: references/format-conventions.md "CREATIVE-DIRECTION.md structure" / "FRAME.md structure".

Full dispatch graphs, spawn mechanics, single-agent fallback, Layer 1→Merge→Layer 2 semantics (coherence check, palette ownership, accessibility hand-back), Step 8.5 7-step projection + invariants, Step 9 sub-paths (9a Paper MCP / 9b Claude Design / 9c None / 9d Brand-Kit Board): references/procedures/dispatch-mechanics.md + references/agent-manifest.md. Step 9a artboard rules: references/artboard-generation.md; Step 9d gated by references/brand-kit-rendering.md + critic BK-G1-G9.

Key invariants (always-on): every ASSETS row has spec ref + target path · ASSETS platform set ≡ declared platforms ≡ BRAND Digital Touchpoints ≡ DESIGN Platform Icon Specs · human [~]/[!] markers never overwritten · no invented rows · Step 8.5 reads references/assets-inventory.md.

Artifact Contract

• Paths (Route B): docs/forsvn/canonical/marketing/{BRAND,DESIGN,CREATIVE-DIRECTION,FRAME,ASSETS}.md (ids brand, design, creative-direction, frame, assets) · Route A: docs/forsvn/canonical/marketing/BRAND.md only.

• Lifecycle: canonical — brand-of-record artifacts consumed by 10+ downstream marketing + product skills.

• Versioning: all overwrite in place + increment the integer version: on re-run (prior versions live in git history). ASSETS.md additionally moves dropped-platform rows to ## Orphaned (preserved). Never a .v[N].md sibling under canonical/ — the UPPERCASE canonical name grammar forbids dots.

• Frontmatter + section schema: references/format-conventions.md.

• Cross-stack contract: schema changes (frontmatter, section headings, table columns) require atomic update of format-conventions.md + every downstream caller (write-copy, write-ad, write-outreach, brief-landing-page, brief-graphic, plan-campaign, humanmaxxing, polish-vn, brief-shortform, map-user-flow). The four review fields are additive — downstream callers consume brand content by heading match.

Full templates: references/artifact-templates.md.

Chain Position

Previous: research-icp (product context) | Next: plan-campaign, write-copy, brief-landing-page, brief-graphic.

Re-run triggers: major product pivots, new markets, audience shifts, annual brand refresh.

Anti-Patterns

Read references/anti-patterns.md before output ships — 21-pattern catalog (13 brand-system pipeline + 4 cross-cutting + 2 narrative-tension + 2 brand-kit board). Most common: aesthetics-without-strategy, generic values, inventing ASSETS rows, overwriting human markers.

Durable Rules (protected)

Completion Status

• DONE — Route A: BRAND.md, critic PASS. Route B: BRAND.md + DESIGN.md + CREATIVE-DIRECTION.md + FRAME.md + ASSETS.md, critic PASS, ASSETS.md auto-scan complete. Refresh-only: the single named artifact (CREATIVE-DIRECTION.md or FRAME.md) rewritten, version bumped, all other canonical files byte-unchanged.
• DONE_WITH_CONCERNS — artifacts written but critic flagged secondary issues (token coverage thin, archetype blend ambiguous, ASSETS.md Orphaned rows surfaced).
• BLOCKED — product context contradictory across audience and positioning (enterprise positioning + consumer voice); needs user reconciliation.
• NEEDS_CONTEXT — no audience/product description AND research/product-context.md absent; recommend research-icp first.

Worked Example

End-to-end Route B (FinLit personal finance app, 3 platforms, full 8-agent dispatch, 74-row ASSETS.md, critic PASS) + Route A (TaskFlow MVP, 2 platforms, reduced critic): references/examples/brand-system-walkthrough.md.