Brandmint

Orchestrated brand identity system that transforms a brand definition into comprehensive marketing outputs — from strategic text documents through AI-generated visual assets to video deliverables and published documentation sites.

Architecture Overview

brand-config.yaml
       ↓
┌─────────────┐     ┌───────────────────┐
│ Wave Planner│←────│Scenario Recommender│
└──────┬──────┘     └───────────────────┘
       ↓
   Waves 1→6
       ↓
   Skills Registry (44 skills, 9 categories)
       ↓
   Hydrator (feeds outputs → config)
       ↓
   Visual Pipeline (FAL.AI assets)
       ↓
   Publishing Pipeline (Wave 7: theme, NotebookLM, decks, reports, diagrams, video)
       ↓
   Wiki → Astro Site

Prerequisites

• Python 3.10+ with uv package manager
• FAL_KEY in ~/.claude/.env
• Dependencies: python-dotenv, fal-client, requests, pyyaml
• Bun (for Astro wiki publishing)
• Node.js >= 18 (for Remotion video rendering)

uv venv .venv && source .venv/bin/activate
uv pip install python-dotenv fal-client requests pyyaml

CLI Commands

# Main entry points (both work)
brandmint [command]
bm [command]

# Global flags (available on all commands)
bm --verbose launch ...   # Enable verbose logging
bm --debug launch ...     # Maximum verbosity
bm --quiet launch ...     # Suppress most output

# Full pipeline wizard
bm launch --config brand-config.yaml --scenario crowdfunding-lean --waves 1-3
bm launch --config brand-config.yaml --max-cost 5.00     # Abort if cost exceeds budget
bm launch --config brand-config.yaml --resume-from 3      # Resume from specific wave

# Visual asset generation (3-phase)
bm visual generate --config brand-config.yaml           # Phase 1: Generate scripts
bm visual execute --config brand-config.yaml --batch anchor     # Phase 2a: Anchor FIRST
bm visual execute --config brand-config.yaml --batch identity   # Phase 2b: Parallel batches
bm visual execute --config brand-config.yaml --batch all        # Or run all remaining
bm visual execute --config brand-config.yaml --force            # Bypass cache, regenerate
bm visual verify --config brand-config.yaml             # Phase 3: Validate

# Execution reports
bm report --config brand-config.yaml                    # Console summary
bm report --config brand-config.yaml --format json      # JSON output
bm report --config brand-config.yaml --format html -o report.html

# Cache management
bm cache stats                                          # Show cache statistics
bm cache clear                                          # Clear all cached prompts
bm cache clear --expired                                # Clear only expired entries

# Scenario planning
bm plan context --config brand-config.yaml
bm plan recommend --config brand-config.yaml

# Skill management
bm install skills    # Creates symlinks in ~/.claude/skills/
bm install check     # Verify installation
bm registry list     # List all 44 skills

Image Providers

Set via env var or config (generation.provider in brand-config.yaml):

Provider	Env Var	Style Anchor	Notes
FAL.AI (default)	FAL_KEY	✅ Full	Best consistency, recommended
OpenRouter	OPENROUTER_API_KEY	❌	Unified API, text-only prompts
OpenAI	OPENAI_API_KEY	⚠️ Limited	DALL-E 3 has fixed sizes
Replicate	REPLICATE_API_TOKEN	⚠️ Limited	Pay-per-second pricing

Note: Only FAL.AI's Nano Banana Pro supports image references (style anchor cascade).

Skill Categories (44 skills)

Category	Count	Purpose
text-strategy/	7	Brand positioning, personas, voice, competitive analysis
visual-prompters/	9	AI image prompt generation (product, fashion, editorial, brand)
campaign-copy/	6	Campaign page copy, ads, hooks, press releases
email-sequences/	3	Welcome, pre-launch, and launch email sequences
brand-foundation/	3	Visual identity, brand naming, logo concept design
social-growth/	5	Social content calendar, community, influencer outreach
advertising/	5	Pre-launch ads, competitive ad extraction, niche validation
publishing/	6	NotebookLM, slide decks, reports, diagrams, video, wiki
visual-pipeline/	4	AI visual asset generation + integration orchestrator

Workflow Routing

• "Create a brand for [X]" → Full orchestration: Waves 1-6 → visual pipeline → publishing
• "Generate assets for [X]" → Execute visual pipeline only (needs existing config)
• "Build wiki from outputs" → Publishing pipeline: wiki-doc-generator → markdown-to-astro-wiki
• "Run [skill-name] for [X]" → Individual skill execution

Wave Execution

Skills execute in dependency-ordered waves. Use depth to control how many waves run:

Depth	Waves	Use Case
surface	1-2	Quick positioning only
focused	1-5	Standard launch (default)
comprehensive	1-6	Full campaign
exhaustive	all	Enterprise/premium

Wave	Skills	Purpose
1	buyer-persona, competitor-analysis	Foundation research
2	product-positioning-summary, mds-messaging-direction-summary	Strategic positioning
3	voice-and-tone, visual-identity-core	Brand personality
4	campaign-page-copy, detailed-product-description	Core copy
5	email sequences, social-content-engine, ads	Channels
6	Visual pipeline, publishing pipeline	Assets & output
7	Theme, NotebookLM, decks, reports, diagrams, video	Deliverables

Scenarios

Pre-built execution profiles that filter skills and set execution context:

Scenario ID	Budget	Best For
brand-genesis	Bootstrapped	Pre-launch foundation
crowdfunding-lean	Lean	Kickstarter/Indiegogo essentials
crowdfunding-full	Standard	Full crowdfunding campaign
bootstrapped-dtc	Bootstrapped	Shopify/organic launch
enterprise-gtm	Premium	B2B SaaS go-to-market
custom-hybrid	Any	Pick-and-choose skills

Domain Tags

Assets are filtered by brand.domain_tags in config. Only matching assets generate:

Tag	Assets Included
* (universal)	2A, 2B (always generated)
dtc, crowdfunding	2C, 3A, 3B, 4A, 4B, 5A-C, 7A, 8A
app, saas	APP-ICON, APP-SCREENSHOT, OG-IMAGE
social	IG-STORY, TWITTER-HEADER
enterprise	PITCH-HERO, 2C

Hydrator

The hydrator feeds text skill outputs back into brand-config.yaml for downstream consumption:

Skill Output	Config Section
buyer-persona	hydrated.buyer_persona
product-positioning-summary	hydrated.positioning
mds-messaging-direction-summary	hydrated.messaging
voice-and-tone	hydrated.voice
competitor-analysis	hydrated.competitors

Backup behavior: save_hydrated_config() creates .yaml.bak before overwriting.

Visual Generation Pipeline

# Phase 1: Generate prompt cookbook + Python scripts
python3 scripts/generate_pipeline.py ./brand-config.yaml

# Phase 2: Execute (anchor first, then parallel batches)
python3 scripts/run_pipeline.py execute --batch anchor
python3 scripts/run_pipeline.py execute --batch identity  # parallel
python3 scripts/run_pipeline.py execute --batch products   # parallel
python3 scripts/run_pipeline.py execute --batch photography # parallel

# Phase 3: Verify
python3 scripts/run_pipeline.py verify --config ./brand-config.yaml

Publishing Pipeline (Wave 7)

# Standalone publishing commands
bm publish notebooklm --config brand-config.yaml    # 7B: NotebookLM notebook + audio
bm publish decks --config brand-config.yaml          # 7C: Slide decks (Marp → PDF)
bm publish reports --config brand-config.yaml        # 7D: Reports (Typst → PDF)
bm publish diagrams --config brand-config.yaml       # 7E: Mind maps & diagrams
bm publish video --config brand-config.yaml          # 7F: Videos (Remotion → MP4)

# Wiki documentation site
./skills/publishing/markdown-to-astro-wiki/scripts/init-astro-wiki.sh my-wiki
./skills/publishing/markdown-to-astro-wiki/scripts/process-markdown.sh ./docs ./my-wiki/src/content/docs
cd my-wiki && bun run build

Key Files

File	Purpose
scripts/generate_pipeline.py	Template engine: config → prompts → generation scripts
scripts/run_pipeline.py	Pipeline executor with batch dispatch
brandmint/core/wave_planner.py	Dependency-ordered skill execution
brandmint/core/hydrator.py	Feeds skill outputs into config
brandmint/core/skills_registry.py	3-source skill discovery
brandmint/core/cache.py	Prompt hash caching for regeneration avoidance
brandmint/cli/report.py	Execution report generator (markdown/json/html)
brandmint/cli/logging.py	Structured logging with Rich integration
brandmint/cli/notifications.py	macOS/Linux desktop + webhook notifications
brandmint/publishing/remotion_generator.py	Remotion video scaffolding + rendering
brandmint/publishing/theme_exporter.py	Brand theme export (CSS/Typst/JSON/Remotion)
brandmint/publishing/marp_generator.py	Marp slide deck generation

Brand Config Schema

Every brand is defined by brand-config.yaml:

• Schema: assets/brand-config-schema.yaml
• Example: assets/example-tryambakam-noesis.yaml
• Key sections: brand, theme, palette, typography, aesthetic, logo_files, products, prompts

Cost Estimation

Item	FAL	OpenRouter	OpenAI
Full brand run (19 assets × 2 seeds)	~$2-3	~$2-2.50	~$3-4
Nano Banana Pro equivalent	$0.08/img	$0.05/img	$0.08/img
Flux 2 Pro equivalent	$0.05/img	$0.05/img	$0.04/img

Use bm visual preview --config brand-config.yaml --json for detailed cost breakdown.

Critical Learnings

1. Anchor cascade — Style anchor bento MUST generate before all other visual assets
2. Recraft V3 returns SVG/WebP — Must detect and convert to PNG
3. Recraft 1000-char limit — Prompts silently truncate
4. Product identity in prompts — Use {product_hero_physical} not generic descriptors
5. Nano Banana aspect ratios — Use 16:9 format not Flux landscape_16_9
6. API keys — Always load_dotenv(os.path.expanduser("~/.claude/.env"))

Trigger Phrases

Claude should invoke this skill when user says:

• "create a brand", "brand identity", "brand launch"
• "generate brand assets", "visual assets for brand"
• "run brandmint", "execute brand workflow"
• "crowdfunding campaign", "launch campaign"
• "brand positioning", "buyer persona"
• "generate visual pipeline", "brand visuals"