AI Agent Teams for Marketers: Build Your Campaign Production System with Skills and Agents

Your next product launch needs a landing page, 5 emails, 3 ad variants, social posts, and tracking documentation. The old way: assign tasks to different team members, wait for handoffs, review drafts, request revisions, repeat for two weeks. The new way: deploy an AI agent system with specialized sub-agents that analyzes your historical data, builds a data-driven GTM plan, and produces all assets in parallel.

This guide shows you how to build a complete campaign production system using Claude’s skills and agents. You get the folder structure, the SKILL.md files, the agent definitions, and the prompts. Works in Claude Code for technical users and as uploadable skills in Claude.ai for everyone else.

TL;DR

AI agent teams use a four-layer architecture: Strategy, Production, Optimization, and Governance. The Production Layer contains sub-agents for each content type: landing pages, emails, blog posts, social content, and ads. Skills encode domain knowledge that agents draw on automatically. You set up the folder structure once and reuse it for every campaign. Time saved: 20-30 hours per campaign launch.

Key Takeaways

• Skills are folders with SKILL.md files that Claude loads automatically when relevant
• Agents are specialized assistants you invoke for specific tasks
• Sub-agents handle individual deliverables within a layer (email writer, landing page writer, blog writer)
• The Production Lead agent spawns sub-agents in parallel for speed
• The GTM Strategy Agent analyzes historical data before production starts
• Governance checks happen after each layer, not before
• This structure works in Claude Code, Cowork, and Claude.ai with uploaded skills

What Is the Complete System Architecture?

The system has four layers. Each layer contains agents. Some agents spawn sub-agents for specific tasks. Skills provide the domain knowledge that agents and sub-agents use.

Two diagrams explain the system. The first shows what components exist. The second shows how data flows when you run a campaign.

Diagram 1: System Components (What You Build)

This diagram shows all the parts and what skills each agent uses. Think of this as your “parts list” for building the system.

SYSTEM COMPONENTS
═══════════════════════════════════════════════════════════════════════════════

ORCHESTRATOR AGENT
File: agents/orchestrator.md
Can spawn: gtm-strategist, production-lead, seo-optimizer, governance-reviewer
│
├── STRATEGY LAYER
│   └── gtm-strategist (agent)
│       File: agents/gtm-strategist.md
│       Uses skills:
│       • campaign-planning
│       • data-analysis
│
├── PRODUCTION LAYER
│   └── production-lead (agent)
│       File: agents/production-lead.md
│       Spawns sub-agents:
│       │
│       ├── landing-page-writer
│       │   File: subagents/landing-page-writer.md
│       │   Uses skills: brand-voice, content-creation, aeo-guidelines
│       │
│       ├── email-writer
│       │   File: subagents/email-writer.md
│       │   Uses skills: brand-voice, content-creation
│       │
│       ├── blog-writer
│       │   File: subagents/blog-writer.md
│       │   Uses skills: brand-voice, content-creation, seo-checklist, aeo-guidelines
│       │
│       ├── social-writer
│       │   File: subagents/social-writer.md
│       │   Uses skills: brand-voice, content-creation
│       │
│       ├── ad-writer
│       │   File: subagents/ad-writer.md
│       │   Uses skills: brand-voice, content-creation, compliance-rules
│       │
│       └── content-validator
│           File: subagents/content-validator.md
│           Uses skills: brand-voice, compliance-rules
│
├── OPTIMIZATION LAYER
│   └── seo-optimizer (agent)
│       File: agents/seo-optimizer.md
│       Uses skills:
│       • seo-checklist
│       • aeo-guidelines
│
└── GOVERNANCE LAYER
    └── governance-reviewer (agent)
        File: agents/governance-reviewer.md
        Uses skills:
        • brand-voice
        • compliance-rules
        • budget-limits

Diagram 2: Execution Flow (How It Runs)

This diagram shows the actual sequence when you run a campaign. Data flows from top to bottom. Each step must complete before the next begins.

EXECUTION FLOW
═══════════════════════════════════════════════════════════════════════════════

                         YOUR INPUTS
               ┌───────────────────────────────┐
               │  • Campaign brief             │
               │  • Historical data export     │
               │  • Brand guidelines           │
               │  • Budget constraints         │
               └───────────────┬───────────────┘
                               │
                               ▼
               ┌───────────────────────────────┐
               │         ORCHESTRATOR          │
               │                               │
               │  Receives your inputs.        │
               │  Coordinates all agents.      │
               └───────────────┬───────────────┘
                               │
                               │ Step 1: "Create GTM Plan"
                               ▼
               ┌───────────────────────────────┐
               │        GTM STRATEGIST         │
               │                               │
               │  Input: Historical data       │
               │  Output: GTM Plan with        │
               │  channels, budget, and        │
               │  deliverables list            │
               └───────────────┬───────────────┘
                               │
                               │ Step 2: "Produce content"
                               ▼
               ┌───────────────────────────────┐
               │       PRODUCTION LEAD         │
               │                               │
               │  Reads GTM Plan               │
               │  Identifies deliverables      │
               │  Spawns sub-agents            │
               └───────────────┬───────────────┘
                               │
                               │ Spawns in PARALLEL
            ┌──────────┬───────┼───────┬──────────┐
            ▼          ▼       ▼       ▼          ▼
         ┌─────┐   ┌─────┐ ┌─────┐ ┌─────┐   ┌─────┐
         │ LP  │   │Email│ │Blog │ │Social│  │ Ads │
         │Writer│  │Writer│ │Writer│ │Writer│ │Writer│
         └──┬──┘   └──┬──┘ └──┬──┘ └──┬──┘   └──┬──┘
            │         │       │       │          │
            │         │       │       │          │
            └─────────┴───┬───┴───────┴──────────┘
                          │
                          │ All content collected
                          ▼
               ┌───────────────────────────────┐
               │      CONTENT VALIDATOR        │
               │                               │
               │  Checks each piece against    │
               │  brand voice and format       │
               │  requirements                 │
               └───────────────┬───────────────┘
                               │
                               │ Step 3: "Optimize content"
                               ▼
               ┌───────────────────────────────┐
               │        SEO OPTIMIZER          │
               │                               │
               │  Input: All validated content │
               │  Output: SEO/AEO scores +     │
               │  optimized versions           │
               └───────────────┬───────────────┘
                               │
                               │ Step 4: "Final checks"
                               ▼
               ┌───────────────────────────────┐
               │     GOVERNANCE REVIEWER       │
               │                               │
               │  Brand compliance check       │
               │  Budget allocation check      │
               │  Legal/regulatory check       │
               └───────────────┬───────────────┘
                               │
                               ▼
               ┌───────────────────────────────┐
               │      LAUNCH-READY ASSETS      │
               │                               │
               │  • GTM Plan                   │
               │  • Optimized landing pages    │
               │  • Email sequences            │
               │  • Blog posts                 │
               │  • Social content             │
               │  • Ad variants                │
               │  • SEO/AEO scores             │
               │  • Governance approval        │
               └───────────────────────────────┘

The key insight: This is a sequential pipeline with one parallel step. Strategy must finish before Production starts. Sub-agents run in parallel for speed. Optimization runs after all content exists. Governance checks the final package.

Action item: Decide which content types you produce for campaigns. Each type becomes a sub-agent in your Production Layer.

What Is the Folder Structure?

Here is the complete folder structure for your marketing agent system. This works in Claude Code projects and as uploaded skills in Claude.ai.

FOLDER STRUCTURE FOR MARKETING AGENT SYSTEM
═══════════════════════════════════════════════════════════════════════════════

your-marketing-project/
│
├── .claude/                          # Claude Code configuration
│   ├── settings.json                 # Project settings
│   │
│   ├── agents/                       # AGENTS: Specialized assistants
│   │   ├── orchestrator.md           # Master coordinator (you talk to this)
│   │   ├── gtm-strategist.md         # Strategy Layer
│   │   ├── production-lead.md        # Production Layer (spawns sub-agents)
│   │   ├── seo-optimizer.md          # Optimization Layer
│   │   └── governance-reviewer.md    # Governance Layer
│   │
│   ├── subagents/                    # SUB-AGENTS: Workers spawned by agents
│   │   ├── landing-page-writer.md    # Creates landing page copy
│   │   ├── email-writer.md           # Creates email sequences
│   │   ├── blog-writer.md            # Creates blog posts
│   │   ├── social-writer.md          # Creates social content
│   │   ├── ad-writer.md              # Creates ad copy
│   │   └── content-validator.md      # Validates all content
│   │
│   └── skills/                       # SKILLS: Domain knowledge (auto-loads)
│       ├── brand-voice/
│       │   └── SKILL.md              # Brand tone, terminology, style
│       ├── content-creation/
│       │   └── SKILL.md              # Writing templates by channel
│       ├── campaign-planning/
│       │   └── SKILL.md              # GTM frameworks, budget allocation
│       ├── seo-checklist/
│       │   └── SKILL.md              # SEO requirements and checks
│       ├── aeo-guidelines/
│       │   └── SKILL.md              # AEO optimization rules
│       └── compliance-rules/
│           └── SKILL.md              # Legal requirements, disclosures
│
├── data/                             # Your campaign data
│   ├── historical-performance.csv    # Export from HubSpot, GA4, etc.
│   ├── brand-guidelines.md           # Your brand document
│   └── campaign-briefs/              # Campaign brief templates
│
└── outputs/                          # Where agents save deliverables
    ├── gtm-plans/
    ├── landing-pages/
    ├── emails/
    ├── blog-posts/
    ├── social/
    └── ads/

Skills load automatically when Claude detects they are relevant. Agents are invoked explicitly or spawned by other agents. Sub-agents are spawned by parent agents to handle specific tasks.

Action item: Create this folder structure in your project directory. Start with the skills folder. Skills are reusable across all campaigns.

What Is the Difference Between Skills, Agents, and Sub-Agents?

These three concepts work together but serve different purposes:

SKILLS vs AGENTS vs SUB-AGENTS
═══════════════════════════════════════════════════════════════════════════════

SKILLS (Auto-Load Domain Knowledge)
───────────────────────────────────
What: Folders with SKILL.md files containing expertise
When: Claude loads them automatically when relevant
Where: .claude/skills/ directory
How: You mention a topic, Claude detects the skill and uses it

Example: You ask about landing page copy.
Claude sees the content-creation skill and loads its guidelines.

File: skills/content-creation/SKILL.md
┌─────────────────────────────────────────────────────────────┐
│  ---                                                         │
│  name: Content Creation                                      │
│  description: Use for any marketing content creation task    │
│  ---                                                         │
│                                                              │
│  ## Headlines                                                │
│  - Under 10 words                                            │
│  - Lead with outcome, not feature                            │
│  - No product name in headline                               │
│                                                              │
│  ## CTAs                                                     │
│  - Start with verb                                           │
│  - Under 5 words                                             │
│  ...                                                         │
└─────────────────────────────────────────────────────────────┘

AGENTS (Specialized Assistants You Invoke)
──────────────────────────────────────────
What: Markdown files defining a specialized role
When: You invoke them explicitly OR orchestrator spawns them
Where: .claude/agents/ directory
How: Type the agent name or let orchestrator coordinate

Example: You say "run the GTM strategist."
Claude loads the gtm-strategist agent and follows its instructions.

File: agents/gtm-strategist.md
┌─────────────────────────────────────────────────────────────┐
│  ---                                                         │
│  description: Creates data-driven go-to-market plans         │
│  capabilities:                                               │
│    - Analyze historical performance data                     │
│    - Recommend channel mix and budget allocation             │
│    - Define success metrics                                  │
│  tools: [read, write, web_search]                            │
│  ---                                                         │
│                                                              │
│  You are a marketing strategist who builds GTM plans.        │
│                                                              │
│  ## Process                                                  │
│  1. Request historical data if not provided                  │
│  2. Analyze patterns in email, landing page, ad performance  │
│  3. Recommend channel mix with data justification            │
│  ...                                                         │
└─────────────────────────────────────────────────────────────┘

SUB-AGENTS (Workers Spawned by Agents)
──────────────────────────────────────
What: Specialized workers for specific deliverables
When: Parent agent spawns them for parallel execution
Where: .claude/subagents/ directory
How: Production-lead agent creates tasks for each sub-agent

Example: Production-lead needs landing page and emails.
It spawns landing-page-writer and email-writer in parallel.

File: subagents/email-writer.md
┌─────────────────────────────────────────────────────────────┐
│  ---                                                         │
│  description: Creates email sequences for campaigns          │
│  capabilities:                                               │
│    - Write welcome sequences                                 │
│    - Write nurture sequences                                 │
│    - Write promotional emails                                │
│  tools: [read, write]                                        │
│  skills_required: [brand-voice, content-creation]            │
│  ---                                                         │
│                                                              │
│  You are an email copywriter specializing in sequences.      │
│                                                              │
│  ## Email Types                                              │
│  - Welcome: Immediate send, 300 words max                    │
│  - Nurture: Days 2-6, 200 words each                         │
│  - Close: Day 8, urgency messaging                           │
│  ...                                                         │
└─────────────────────────────────────────────────────────────┘

The hierarchy: Orchestrator spawns Agents. Agents spawn Sub-agents. All of them use Skills for domain knowledge.

How Does the Production Lead Spawn Sub-Agents?

The Production Lead agent is the manager of content creation. It reads the GTM Plan, identifies required deliverables, and spawns the right sub-agents in parallel.

PRODUCTION LEAD WORKFLOW
═══════════════════════════════════════════════════════════════════════════════

INPUT: GTM Plan from Strategy Layer
─────────────────────────────────────
GTM Plan says:
• Landing pages: 2 (main + retargeting)
• Email sequence: 5 emails
• Blog post: 1 (SEO-focused)
• Social posts: 8 (2 per platform)
• Google ads: 6 variants
• LinkedIn ads: 4 variants

PRODUCTION LEAD READS GTM PLAN
──────────────────────────────
┌─────────────────────────────────────────────────────────────────────────────┐
│  Production Lead agent:                                                      │
│                                                                              │
│  1. Parses deliverable list from GTM Plan                                    │
│  2. Maps each deliverable to a sub-agent                                     │
│  3. Creates task for each sub-agent with specific instructions               │
│  4. Spawns all sub-agents IN PARALLEL                                        │
│  5. Waits for all to complete                                                │
│  6. Runs validator sub-agent on each output                                  │
│  7. Collects validated outputs                                               │
│  8. Passes to Optimization Layer                                             │
└─────────────────────────────────────────────────────────────────────────────┘
                                   │
                                   │ Spawns in parallel:
                                   │
         ┌─────────────────────────┼─────────────────────────┐
         │                         │                         │
         ▼                         ▼                         ▼
┌─────────────────┐      ┌─────────────────┐      ┌─────────────────┐
│ LANDING PAGE    │      │ EMAIL WRITER    │      │ BLOG WRITER     │
│ WRITER          │      │                 │      │                 │
│                 │      │ Task: 5 emails  │      │ Task: 1 post    │
│ Task: 2 pages   │      │ • Welcome       │      │ • SEO-focused   │
│ • Main LP       │      │ • Value x2      │      │ • 1500 words    │
│ • Retargeting   │      │ • Objection     │      │ • AEO optimized │
│                 │      │ • Close         │      │                 │
│ Skills used:    │      │                 │      │ Skills used:    │
│ • brand-voice   │      │ Skills used:    │      │ • brand-voice   │
│ • content-create│      │ • brand-voice   │      │ • content-create│
│ • aeo-guidelines│      │ • content-create│      │ • seo-checklist │
└────────┬────────┘      └────────┬────────┘      │ • aeo-guidelines│
         │                        │               └────────┬────────┘
         │                        │                        │
         ▼                        ▼                        ▼
┌─────────────────┐      ┌─────────────────┐      ┌─────────────────┐
│ SOCIAL WRITER   │      │ AD WRITER       │      │                 │
│                 │      │                 │      │                 │
│ Task: 8 posts   │      │ Task: 10 ads    │      │                 │
│ • LinkedIn x2   │      │ • Google x6     │      │                 │
│ • Twitter x2    │      │ • LinkedIn x4   │      │                 │
│ • Facebook x2   │      │                 │      │                 │
│ • Instagram x2  │      │ Skills used:    │      │                 │
│                 │      │ • brand-voice   │      │                 │
│ Skills used:    │      │ • content-create│      │                 │
│ • brand-voice   │      │ • compliance    │      │                 │
│ • content-create│      │                 │      │                 │
└────────┬────────┘      └────────┬────────┘      │                 │
         │                        │               │                 │
         └────────────────────────┼───────────────┘                 │
                                  │                                 │
                                  ▼                                 │
                    ┌─────────────────────────────┐                 │
                    │      CONTENT VALIDATOR      │◄────────────────┘
                    │                             │
                    │  Runs after ALL sub-agents  │
                    │  complete. Checks each      │
                    │  output against:            │
                    │  • Brand voice guidelines   │
                    │  • Character/word limits    │
                    │  • Required elements        │
                    │  • Compliance rules         │
                    └─────────────────────────────┘
                                  │
                                  ▼
                    All validated content passed to
                    SEO/AEO Optimizer

All sub-agents run at the same time. The landing page writer does not wait for the email writer. This parallel execution cuts production time from days to hours.

Action item: List every content type your team produces. Create a sub-agent file for each. Even if you start with just one, the structure scales.

What Do the SKILL.md Files Look Like?

Skills are simple markdown files with YAML frontmatter. Here are the essential skills for a marketing agent system:

Brand Voice Skill

FILE: .claude/skills/brand-voice/SKILL.md
═══════════════════════════════════════════════════════════════════════════════

---
name: Brand Voice
description: Use when creating any customer-facing content. Defines tone, terminology, and style guidelines.
version: 1.0.0
---

# Brand Voice Skill

## Tone Attributes

Write as a confident practitioner, not a thought leader. Direct language. Short sentences. No fluff.

## Voice Guidelines

Use active voice always. Lead with the outcome. Address the reader as "you." State facts, do not hedge.

## Banned Words

Never use: delve, landscape, game-changer, unlock, revolutionize, leverage, synergy, robust, seamless, cutting-edge

## Terminology

Use "campaign" not "initiative." Use "metrics" not "KPIs" unless in executive context. Use "customer" not "user" in external content.

## Sentence Structure

Maximum 20 words per sentence. One idea per sentence. No semicolons in customer-facing copy.

Content Creation Skill

FILE: .claude/skills/content-creation/SKILL.md
═══════════════════════════════════════════════════════════════════════════════

---
name: Content Creation
description: Use for any marketing content creation. Contains templates and best practices by channel.
version: 1.0.0
---

# Content Creation Skill

## Landing Page Structure

Headline: Under 10 words. Lead with outcome, not product name.
Subheadline: One sentence expanding the benefit.
Hero copy: 2-3 sentences addressing pain point.
Benefits: 4-6 bullets starting with verbs.
Social proof: Testimonial with name, title, company.
CTA button: Under 5 words, starts with verb.
Supporting text: One sentence reducing friction.

## Email Structure

Subject line: Under 50 characters. No spam trigger words.
Preview text: Complements subject, does not repeat it.
Opening: Personal, references previous interaction or pain point.
Body: Under 200 words for nurture, 300 for welcome.
CTA: One primary action per email.
PS line: Optional, use for urgency or social proof.

## Blog Post Structure

Title: Under 60 characters. Front-load primary keyword.
Meta description: 155-160 characters.
H1: Matches title or close variant.
Opening: State the core value in first paragraph.
H2s: Question-style for AEO optimization.
Paragraphs: Under 80 words each.
CTAs: One per major section.

Campaign Planning Skill

FILE: .claude/skills/campaign-planning/SKILL.md
═══════════════════════════════════════════════════════════════════════════════

---
name: Campaign Planning
description: Use when creating GTM plans, allocating budgets, or selecting channels.
version: 1.0.0
---

# Campaign Planning Skill

## Channel Selection Framework

Evaluate channels based on historical data:

1. Calculate CPA by channel for similar campaigns
2. Calculate ROAS by channel for similar offers
3. Identify audience overlap with target segment
4. Assess content requirements and production cost
5. Rank channels by efficiency (ROAS / production effort)

## Budget Allocation Rules

Primary channel: 40-50% of budget (highest historical ROAS)
Secondary channel: 25-35% (second-highest ROAS or strategic bet)
Tertiary channel: 15-25% (testing or audience expansion)
Buffer: Always reserve 10% for optimization

## Success Metrics by Channel

Email: Open rate, click rate, conversion rate, revenue per send
Paid search: CTR, CPC, CPA, ROAS, impression share
Paid social: CTR, CPC, CPM, frequency, relevance score
Organic: Rankings, traffic, time on page, conversion rate

SEO Checklist Skill

FILE: .claude/skills/seo-checklist/SKILL.md
═══════════════════════════════════════════════════════════════════════════════

---
name: SEO Checklist
description: Use when optimizing content for search engines. Contains technical and on-page requirements.
version: 1.0.0
---

# SEO Checklist Skill

## On-Page Requirements

Title tag: 50-60 characters, primary keyword in first 60 chars
Meta description: 155-160 characters, includes CTA
H1: One only, contains primary keyword
H2s: 3-7 per page, include keyword variations
Keyword density: 1-2% for primary keyword
First 100 words: Primary keyword present
Internal links: 2-3 minimum to related pages
External links: 1-2 to authoritative sources
Image alt text: Descriptive, keyword where natural
URL: Short, includes primary keyword, no dates

## Technical Requirements

HTTPS: Required
Mobile responsive: Required
Page speed: Under 3 seconds LCP
Schema markup: Article, FAQ, or HowTo as appropriate
Canonical URL: Set to prevent duplicate content

AEO Guidelines Skill

FILE: .claude/skills/aeo-guidelines/SKILL.md
═══════════════════════════════════════════════════════════════════════════════

---
name: AEO Guidelines
description: Use when optimizing content for AI assistants like ChatGPT, Claude, and Perplexity.
version: 1.0.0
---

# AEO Guidelines Skill

## Paragraph Rules

Length: Under 80 words per paragraph (AI quotes short text)
First sentence: Contains the direct answer (AI reads this first)
Structure: One idea per paragraph only
No fluff: Cut "In today's world" and similar openings

## Heading Rules

Format: Question-style matching user queries
Examples: "What Benefits Does X Provide?" not "Benefits"
Placement: H2 for main questions, H3 for follow-ups

## Answer Format

Front-load data: Statistics in first sentence, not buried
Self-contained: Each paragraph should stand alone as a quote
Structured: Use lists and tables for comparison data
Quotable: Write statements AI assistants can cite directly

## Testing

Paste a paragraph into ChatGPT or Claude. Ask it to answer the implied question. If it quotes your paragraph back, it passes.

Action item: Create your brand-voice SKILL.md first. This skill applies to every content type. Get your banned words and terminology documented.

What Do the Agent Files Look Like?

Agents define specialized roles with capabilities and instructions. Here are the core agents:

Orchestrator Agent

FILE: .claude/agents/orchestrator.md
═══════════════════════════════════════════════════════════════════════════════

---
description: Master coordinator for campaign production. Manages all other agents.
capabilities:
  - Coordinate multi-agent workflows
  - Route tasks to appropriate agents
  - Ensure correct execution sequence
tools: [read, write, spawn_agent]
can_spawn: [gtm-strategist, production-lead, seo-optimizer, governance-reviewer]
---

You are the campaign orchestrator. You coordinate AI agent teams to produce marketing campaigns.

## Your Role

You receive campaign briefs and historical data from the user. You manage the workflow through four phases: Strategy, Production, Optimization, and Governance.

## Execution Sequence

Phase 1 - STRATEGY: Spawn gtm-strategist agent. Wait for GTM Plan. Do not proceed until complete.

Phase 2 - PRODUCTION: Pass GTM Plan to production-lead agent. Wait for all content. Do not proceed until complete.

Phase 3 - OPTIMIZATION: Pass content to seo-optimizer agent. Wait for optimization report. Do not proceed until complete.

Phase 4 - GOVERNANCE: Pass all outputs to governance-reviewer agent. Wait for approval.

## Rules

Never skip phases. Each phase depends on the previous. Always confirm completion before proceeding. If any agent reports failure, pause and report to user.

GTM Strategist Agent

FILE: .claude/agents/gtm-strategist.md
═══════════════════════════════════════════════════════════════════════════════

---
description: Creates data-driven go-to-market plans from historical performance data.
capabilities:
  - Analyze historical campaign performance
  - Recommend channel mix with data justification
  - Allocate budget based on efficiency metrics
  - Define success metrics and targets
tools: [read, write, web_search]
skills_required: [campaign-planning, data-analysis]
---

You are a marketing strategist who builds GTM plans based on data.

## Input Requirements

You need historical performance data to make recommendations. If not provided, request:

- Email metrics: Open rates, click rates, conversion rates by segment
- Landing page metrics: Conversion rates, bounce rates by source
- Ad metrics: CTR, CPA, ROAS by platform and audience
- Previous campaign results with spend and outcomes

## Output Format

Your GTM Plan must include:

1. DATA ANALYSIS SUMMARY: Top 3 insights from historical performance
2. CHANNEL MIX: Primary/secondary/tertiary with data justification
3. BUDGET ALLOCATION: Specific amounts totaling the budget
4. CONTENT REQUIREMENTS: Exact deliverables per channel
5. TIMELINE: Week-by-week breakdown
6. SUCCESS METRICS: Targets based on historical benchmarks

## Rules

Never recommend channels without data justification. Never allocate budget without efficiency rationale. Always include a 10% buffer.

Production Lead Agent

FILE: .claude/agents/production-lead.md
═══════════════════════════════════════════════════════════════════════════════

---
description: Manages content production by spawning specialized sub-agents.
capabilities:
  - Parse deliverables from GTM Plan
  - Spawn content sub-agents in parallel
  - Coordinate validation of all outputs
  - Compile final content package
tools: [read, write, spawn_subagent]
can_spawn: [landing-page-writer, email-writer, blog-writer, social-writer, ad-writer, content-validator]
skills_required: [content-creation, brand-voice]
---

You are the production lead. You manage a team of content sub-agents.

## Process

1. Receive GTM Plan from orchestrator
2. Parse the CONTENT REQUIREMENTS section
3. Map each deliverable to a sub-agent
4. Spawn all relevant sub-agents IN PARALLEL
5. Wait for all to complete
6. Spawn content-validator to check each output
7. Collect validated outputs
8. Return complete package to orchestrator

## Sub-Agent Mapping

Landing pages → landing-page-writer
Email sequences → email-writer
Blog posts → blog-writer
Social posts → social-writer
Ad copy → ad-writer

## Rules

Spawn sub-agents in parallel for speed. Do not wait for one to finish before starting another. Run validator after ALL content is complete, not after each piece.

SEO Optimizer Agent

FILE: .claude/agents/seo-optimizer.md
═══════════════════════════════════════════════════════════════════════════════

---
description: Optimizes content for search engines and AI assistants.
capabilities:
  - Audit content for SEO compliance
  - Audit content for AEO compliance
  - Provide specific fixes with before/after text
  - Score content on 0-100 scale
tools: [read, write]
skills_required: [seo-checklist, aeo-guidelines]
---

You are an SEO and AEO optimization specialist.

## Analysis Process

For each piece of content:

1. Run SEO checklist (from seo-checklist skill)
2. Run AEO checklist (from aeo-guidelines skill)
3. Score each criterion as PASS or FAIL
4. Calculate overall scores (0-100)
5. List critical issues with exact fixes
6. Provide optimized content with all fixes applied

## Output Format

SEO Score: X/100
AEO Score: X/100

CRITICAL ISSUES:
1. [Issue]: [Current text] → [Fixed text]
2. [Issue]: [Current text] → [Fixed text]

OPTIMIZED CONTENT:
[Complete revised content]

## Rules

Never approve paragraphs over 80 words. Never approve H2s that don't match search intent. Always provide exact revised text, not suggestions.

Action item: Start with the orchestrator agent. Test it with a simple campaign brief. Add other agents as you validate the workflow.

What Do the Sub-Agent Files Look Like?

Sub-agents are specialists spawned by the Production Lead. Here are the key sub-agents:

Landing Page Writer Sub-Agent

FILE: .claude/subagents/landing-page-writer.md
═══════════════════════════════════════════════════════════════════════════════

---
description: Creates landing page copy for campaigns.
capabilities:
  - Write headlines and subheadlines
  - Write hero copy and benefits
  - Write CTAs and supporting text
tools: [read, write]
skills_required: [brand-voice, content-creation, aeo-guidelines]
---

You are a conversion-focused landing page copywriter.

## Output Structure

For each landing page, provide:

HEADLINE: Under 10 words. Lead with outcome.
SUBHEADLINE: One sentence expanding the benefit.
HERO COPY: 2-3 sentences on pain point.
BENEFITS: 4-6 bullets starting with verbs.
SOCIAL PROOF: Testimonial placeholder with format.
PRIMARY CTA: Button text (under 5 words) + supporting text.
SECONDARY CTA: Alternative action for non-ready visitors.

## Rules

No product name in headline. All paragraphs under 80 words. Benefits focus on outcomes, not features.

Email Writer Sub-Agent

FILE: .claude/subagents/email-writer.md
═══════════════════════════════════════════════════════════════════════════════

---
description: Creates email sequences for campaigns.
capabilities:
  - Write welcome emails
  - Write nurture sequences
  - Write promotional emails
tools: [read, write]
skills_required: [brand-voice, content-creation]
---

You are an email copywriter specializing in sequences.

## Email Types and Structure

WELCOME (Day 0): 300 words max, set expectations, one clear CTA
VALUE (Days 2,4): 200 words max, one key insight each
OBJECTION (Day 6): Address top concern, risk reversal messaging
CLOSE (Day 8): Urgency or scarcity, final push

## For Each Email Provide

Subject line: Under 50 characters
Preview text: Complements subject
Opening hook: Personal, relevant
Body copy: Under word limit
CTA button: Action-oriented
PS line: Optional, for urgency or proof

## Rules

No spam trigger words in subjects. One CTA per email. Include {{FIRST_NAME}} personalization token.

Blog Writer Sub-Agent

FILE: .claude/subagents/blog-writer.md
═══════════════════════════════════════════════════════════════════════════════

---
description: Creates SEO and AEO optimized blog posts.
capabilities:
  - Write long-form blog content
  - Optimize for search engines
  - Optimize for AI assistants
tools: [read, write, web_search]
skills_required: [brand-voice, content-creation, seo-checklist, aeo-guidelines]
---

You are a content writer specializing in SEO and AEO optimized blog posts.

## Output Structure

TITLE: Under 60 characters, primary keyword front-loaded
META DESCRIPTION: 155-160 characters
H1: Matches or close to title
OPENING: Core value in first paragraph
BODY: H2 sections with question-style headings
PARAGRAPHS: Under 80 words each, one idea each
CTAs: One action item per major section
CONCLUSION: Summary without "in conclusion" language

## Rules

All H2s phrased as questions matching search intent. Primary keyword in first 10 words of each section. Front-load data and statistics in first sentence.

Content Validator Sub-Agent

FILE: .claude/subagents/content-validator.md
═══════════════════════════════════════════════════════════════════════════════

---
description: Validates all content against brand and quality standards.
capabilities:
  - Check content against brand guidelines
  - Check content against format requirements
  - Identify issues and provide fixes
tools: [read]
skills_required: [brand-voice, compliance-rules]
---

You are a quality reviewer for marketing content.

## Validation Process

For each piece of content, check:

BRAND: Tone matches guidelines, no banned words, terminology correct
FORMAT: Word counts within limits, required sections present
COMPLIANCE: Legal disclosures present, claims substantiated
QUALITY: No typos, grammar correct, clear messaging

## Output Format

Status: APPROVED or NEEDS REVISION
Checklist: PASS/FAIL for each criterion
Issues: Quote exact text that fails
Fixes: Provide corrected text

## Rules

Quote exact failing text. Provide exact corrected text. Do not approve content with critical failures.

Action item: Create the email-writer sub-agent first. Email sequences are the most common campaign deliverable. Test it before adding other sub-agents.

Which Platform Should You Use?

Marketers have three options for running this system. Each has different capabilities. This section clarifies what works where and where your output files end up.

Platform Comparison Table

FEATURE COMPARISON: CLAUDE.AI vs CLAUDE CODE vs COWORK
═══════════════════════════════════════════════════════════════════════════════

                          │ Claude.ai      │ Claude Code    │ Cowork
                          │ (Pro/Max/Team) │ (CLI or Desktop)│ (Desktop App)
──────────────────────────┼────────────────┼────────────────┼────────────────
Skills (SKILL.md)         │ ✅ Settings >  │ ✅ .claude/    │ ✅ Via plugins
                          │    Capabilities│    skills/     │
──────────────────────────┼────────────────┼────────────────┼────────────────
Skills invoke skills      │ ✅ Sequential  │ ✅ Sequential  │ ✅ Sequential
(orchestration)           │    in convo    │    or parallel │    or parallel
──────────────────────────┼────────────────┼────────────────┼────────────────
True parallel sub-agents  │ ❌ Sequential  │ ✅ Yes         │ ✅ Yes
                          │    only        │                │
──────────────────────────┼────────────────┼────────────────┼────────────────
Slash Commands            │ ❌ No          │ ✅ /commands/  │ ✅ /commands/
(/draft-content)          │                │                │
──────────────────────────┼────────────────┼────────────────┼────────────────
MCP Connectors            │ ✅ Remote MCP  │ ✅ Local +     │ ✅ Local +
(Asana, Notion, etc.)     │    servers     │    Remote      │    Remote
──────────────────────────┼────────────────┼────────────────┼────────────────
File creation             │ ✅ Download    │ ✅ Saves to    │ ✅ Saves to
                          │    to computer │    project     │    local folder
──────────────────────────┼────────────────┼────────────────┼────────────────
Persistent folder         │ ❌ You organize│ ✅ Yes         │ ✅ Yes
structure                 │    downloads   │                │
──────────────────────────┼────────────────┼────────────────┼────────────────
Interface options         │ Web, Mobile,   │ Terminal CLI   │ Desktop GUI
                          │ Desktop app    │ or Desktop GUI │    (task-focused)
──────────────────────────┼────────────────┼────────────────┼────────────────
Requires terminal         │ No             │ No (Desktop    │ No
                          │                │    available)  │
──────────────────────────┼────────────────┼────────────────┼────────────────
Best for                  │ Solo marketers │ All skill      │ Marketing
                          │                │    levels      │    teams

Option 1: Claude.ai Pro/Max/Team (Skills + Orchestration, No Parallel Execution)

Use this if you want skills-based orchestration without terminal commands. Skills CAN invoke other skills in sequence. You get MCP connectors for tools like Asana, Notion, and Linear. Files download to your computer.

What works:

• Upload custom skills via Settings > Capabilities > Skills (as .zip)
• Skills auto-activate when relevant to your request
• Orchestrator skills invoke other skills sequentially
• MCP connectors to Asana, Notion, Linear, Zapier, Jira, Confluence, Slack (Team/Enterprise), and more
• Pre-built skills for docx, pptx, xlsx, pdf work automatically
• File creation with download to your computer

What does not work:

• No true parallel sub-agent execution (skills run sequentially)
• No slash commands
• No persistent folder structure on server (you organize downloads)
• Files do not save to a project folder; you download each one

Where outputs go:

• Claude creates files in a sandboxed environment
• You download each file to your Downloads folder
• You organize files locally after downloading

How orchestration works in Claude.ai:

Your campaign-orchestrator skill example shows the pattern. The orchestrator skill references other skills by name. Claude executes them in sequence within the same conversation.

CLAUDE.AI SKILL ORCHESTRATION FLOW
═══════════════════════════════════════════════════════════════════════════════

Phase 0: Setup
──────────────
User: "Run a full campaign for product X"
Claude: Activates campaign-orchestrator skill
        Creates folder structure references
        Asks for project slug and depth preference

Phase 1: Strategy (Sequential)
──────────────────────────────
Orchestrator invokes: launch-strategy skill
        → Creates GTM plan
        → Invokes cai-brand-guidelines for voice check
        → Invokes competitor-alternatives for analysis
        → Invokes analytics-tracking for measurement plan
        
Claude: "Phase 1 complete. Here are the outputs. Approve to continue."
User: "Approved"

Phase 2: Messaging (Sequential)
───────────────────────────────
Orchestrator invokes: copywriting skill
        → Creates core messaging document
        → References brand guidelines
        → Creates journey map
        
Claude: "Phase 2 complete. Approve to continue."

Phase 3: Channel Assets (Sequential)
────────────────────────────────────
Orchestrator invokes: paid-ads skill → ad copy
Orchestrator invokes: email-sequence skill → emails
Orchestrator invokes: social-content skill → social calendar

Each skill runs one at a time. Not parallel.

Phase 4: Web (Sequential)
─────────────────────────
Orchestrator invokes: page-cro skill → landing page spec

Output: Multiple downloadable files

Key limitation: Skills run sequentially in the same conversation. If you have 5 content types to create, Claude creates them one after another. This takes longer than Claude Code or Cowork where sub-agents run in parallel.

Folder structure (on your computer after downloading):

your-marketing-folder/
├── skills/                    # Your skill .zip files (for uploading)
│   ├── campaign-orchestrator.zip
│   ├── brand-voice.zip
│   ├── content-creation.zip
│   └── seo-checklist.zip
│
├── campaigns/                 # Organized downloads
│   └── 2026-q1-launch/
│       ├── 00_strategy/
│       │   ├── gtm_plan.md
│       │   └── measurement_plan.md
│       ├── 01_messaging/
│       │   └── core_message.md
│       ├── 02_channels/
│       │   ├── paid_ads.md
│       │   └── email_sequence.md
│       └── 03_web/
│           └── landing_page_spec.md

MCP Connectors available in Claude.ai:

Connect these tools via Settings > Connectors or the + button in chat:

• Asana (project management)
• Notion (docs and databases)
• Linear (issue tracking)
• Jira and Confluence (Atlassian)
• Zapier (automation)
• Slack (Team/Enterprise plans)
• Intercom (customer support)
• Google Drive and Gmail
• And more via remote MCP servers

Action item: Upload an orchestrator skill to Claude.ai. Connect one MCP tool you use daily. Run a campaign and see skills invoke skills in sequence.

Option 2: Claude Code (CLI or Desktop, True Parallel Execution)

Use this if you want true parallel sub-agent execution and persistent file organization. Available as terminal CLI or desktop app. Sub-agents run simultaneously. Outputs save to your project folder automatically.

What works:

• Full skill, agent, and sub-agent system
• True parallel execution: 5 sub-agents create content simultaneously
• Slash commands (/draft-content, /seo-audit)
• MCP connectors (local and remote)
• Files save to project folder automatically
• Persistent folder structure
• Desktop GUI available (no terminal required)

What you need:

• Claude Code CLI (terminal) OR Claude Code Desktop (GUI)
• Project folder with .claude/ structure
• Plugins installed via CLI or desktop app

Where outputs go:

your-project/
├── .claude/
│   ├── agents/
│   ├── subagents/
│   └── skills/
│
├── data/
│   └── historical-performance.csv
│
└── outputs/                   # ← Agents save files here automatically
    ├── gtm-plans/
    │   └── 2026-02-launch-gtm.md
    ├── landing-pages/
    │   ├── main-lp.md
    │   └── retargeting-lp.md
    ├── emails/
    │   ├── email-1-welcome.md
    │   ├── email-2-value.md
    │   └── ...
    ├── blog-posts/
    │   └── seo-optimized-post.md
    ├── social/
    │   └── social-variants.md
    └── ads/
        ├── google-ads.md
        └── linkedin-ads.md

How parallel execution works:

CLAUDE CODE PARALLEL EXECUTION
═══════════════════════════════════════════════════════════════════════════════

Step 1: Orchestrator receives brief
───────────────────────────────────
You: "Launch campaign for product X. Data: @data/performance.csv"

Step 2: GTM Strategist runs (sequential, needs to complete first)
─────────────────────────────────────────────────────────────────
Orchestrator spawns GTM Strategist agent
GTM Strategist creates GTM Plan
Saves to outputs/gtm-plans/

Step 3: Production Lead spawns sub-agents (PARALLEL)
────────────────────────────────────────────────────
Production Lead reads GTM Plan
Spawns 5 sub-agents SIMULTANEOUSLY:

    ┌─────────────────┐
    │ Landing Page    │──→ outputs/landing-pages/main-lp.md
    │ Writer          │
    └─────────────────┘
              │
    ┌─────────────────┐
    │ Email Writer    │──→ outputs/emails/*.md
    └─────────────────┘
              │           ALL RUN AT THE SAME TIME
    ┌─────────────────┐
    │ Blog Writer     │──→ outputs/blog-posts/*.md
    └─────────────────┘
              │
    ┌─────────────────┐
    │ Social Writer   │──→ outputs/social/*.md
    └─────────────────┘
              │
    ┌─────────────────┐
    │ Ad Writer       │──→ outputs/ads/*.md
    └─────────────────┘

All 5 complete in parallel. Total time = slowest sub-agent, not sum of all.

Step 4: SEO Optimizer runs (sequential, needs all content)
──────────────────────────────────────────────────────────
Reads all content from outputs/
Optimizes each piece
Updates files in place

Step 5: Governance check (sequential)
─────────────────────────────────────
Reviews all content
Approves or flags issues

TIME: 30-60 minutes (parallel execution saves 2-3 hours)

Workflow:

CLAUDE CODE WORKFLOW
═══════════════════════════════════════════════════════════════════════════════

Step 1: Set up project (one-time)
─────────────────────────────────
mkdir my-campaign && cd my-campaign
# Create .claude/ folder structure
# Add agent, subagent, and skill files

Step 2: Install marketing plugin (optional)
───────────────────────────────────────────
claude plugins add knowledge-work-plugins/marketing

Step 3: Run campaign
────────────────────
claude

You: "Launch a campaign for [product]. Here's my historical data: @data/performance.csv
      Use the orchestrator to coordinate GTM planning and content production."

Claude: [Orchestrator agent activates]
        [Spawns GTM Strategist → waits for GTM Plan]
        [Spawns Production Lead → spawns 5 sub-agents in parallel]
        [Collects all content → runs SEO Optimizer]
        [Runs Governance check → outputs to /outputs/]

"Campaign assets ready. 12 files saved to /outputs/. Here's the summary..."

Step 4: Review outputs
──────────────────────
All files in outputs/ folder. Open in VS Code or your editor.

Action item: If you want parallel execution, install Claude Code. Copy the .claude/ folder structure from this guide. Run your first campaign with parallel sub-agents.

Option 3: Cowork (Full System with GUI, True Parallel Execution)

Use this if you want Claude Code capabilities without terminal commands. Cowork is Claude Code with a graphical interface. Same parallel execution. Same persistent file organization. Friendlier for non-technical users.

What works:

• Everything Claude Code does
• True parallel sub-agent execution
• Visual interface instead of terminal
• Plugin installation via app menu
• File browser for outputs
• Better for teams (easier onboarding)

What you need:

• Claude Desktop app (macOS currently, Windows coming)
• Cowork mode enabled
• Marketing plugin installed via app

Where outputs go:

You choose the output folder in Cowork settings.

Typical setup:
~/Documents/Campaigns/
├── 2026-02-product-launch/
│   ├── 00_strategy/
│   │   └── gtm-plan.md
│   ├── 01_messaging/
│   │   └── core-message.md
│   ├── 02_channels/
│   │   ├── emails/
│   │   ├── ads/
│   │   └── social/
│   └── 03_web/
│       └── landing-page-spec.md
│
└── 2026-03-webinar-promo/
    └── ...

How plugins work in Cowork:

Plugins bundle skills, connectors, commands, and sub-agents into one package. Anthropic released 11 open-source plugins on January 30, 2026:

• Productivity (tasks, calendars, workflows)
• Enterprise search (find docs across tools)
• Sales (prospect research, call prep)
• Finance (financial analysis, models)
• Data (queries, visualizations)
• Legal (contract review, research)
• Marketing (campaigns, content)
• Customer support (ticket handling)
• Project management (specs, roadmaps)
• Biology research (specialized)
• Plugin Create (build your own)

Workflow:

COWORK GUI WORKFLOW
═══════════════════════════════════════════════════════════════════════════════

Step 1: Install marketing plugin
────────────────────────────────
Open Claude Desktop app
Click "Cowork" tab
Go to Plugins > Browse
Find "marketing" plugin
Click Install

Step 2: Configure your brand (one-time)
───────────────────────────────────────
Marketing plugin prompts you to set up:
- Brand voice settings
- Style guide references
- Target personas
These persist across sessions.

Step 3: Run campaign
────────────────────
Switch to Cowork mode (Tasks tab)
Type: "Launch a full campaign for [product]"
Or use slash command: /marketing:campaign-plan

Cowork shows you the plan before executing.
Click "Let's go" to start.
Progress shows in the interface.
Sub-agents run in parallel (you see multiple progress indicators).

Step 4: Review outputs
──────────────────────
Files appear in the output folder you specified.
Use Cowork's file browser to preview.
Export to Google Drive, Notion, CMS.

TIME: 30-60 minutes (same as Claude Code, parallel execution)
EFFORT: Low - GUI makes everything visible

Cowork vs Claude Code:

Aspect	Claude Code	Cowork
Interface	Terminal	GUI
Setup	Command line	Click to install
Parallel execution	Yes	Yes
Plugins	Same format	Same format
Output location	Project folder	Folder you choose
Best for	Technical users	Non-technical users

Action item: Download Claude Desktop. Enable Cowork. Install the marketing plugin. You get parallel sub-agent execution with a visual interface.

Quick Decision Guide

WHICH PLATFORM SHOULD YOU USE?
═══════════════════════════════════════════════════════════════════════════════

START HERE
    │
    ▼
Do you need TRUE PARALLEL sub-agent execution?
(5 content pieces created simultaneously, not sequentially)
    │
    ├── NO → Use Claude.ai Pro/Max/Team
    │        Skills invoke skills sequentially
    │        MCP connectors work
    │        Files download to your computer
    │        Good for: Solo marketers, smaller campaigns
    │
    └── YES → Choose your interface:
                  │
                  ├── TERMINAL → Claude Code CLI
                  │              Command line interface
                  │              Maximum customization
                  │              Good for: Developers, custom workflows
                  │
                  ├── DESKTOP APP → Claude Code Desktop or Cowork
                  │                 Visual interface
                  │                 Plugin installation via GUI
                  │                 Good for: Marketers who want parallel
                  │                           execution without terminal
                  │
                  └── EITHER → Both support the same plugins
                              Same .claude/ folder structure
                              Same parallel sub-agent execution
                              Same file persistence

All three paths to parallel execution:

1. Claude Code CLI - Terminal commands. Install plugins via claude plugins add.
2. Claude Code Desktop - GUI app. Install plugins via menu.
3. Cowork - GUI app with task-focused interface. Install plugins via menu.

Claude Code Desktop and Cowork both run the same underlying system. The difference is interface style:

• Claude Code Desktop: Developer-focused, shows more technical detail
• Cowork: Task-focused, shows progress and outcomes

The Real Difference:

Capability	Claude.ai	Claude Code / Cowork
Skills	✅	✅
Skills invoke skills	✅ Sequential	✅ Sequential or Parallel
Sub-agents	❌	✅ True parallel
MCP Connectors	✅ Remote MCP	✅ Local + Remote
File persistence	❌ Download each	✅ Auto-save to folder
Slash commands	❌	✅
Requires terminal	N/A	❌ Not anymore

Skills work across all platforms. The key differences are parallel execution and file persistence. If you create 5 content types:

• Claude.ai: Creates them one at a time. You download each file.
• Claude Code/Cowork: Creates all 5 simultaneously. Saves all to your project folder.

For a 5-asset campaign, parallel execution saves 60-90 minutes.

What Does the Complete Workflow Look Like?

Here is the step-by-step workflow from brief to launch-ready assets:

COMPLETE CAMPAIGN PRODUCTION WORKFLOW
═══════════════════════════════════════════════════════════════════════════════

SETUP (One-Time)
────────────────
□ Create folder structure
□ Write brand-voice SKILL.md
□ Write content-creation SKILL.md
□ Write seo-checklist SKILL.md
□ Write aeo-guidelines SKILL.md
□ Create agent files
□ Create sub-agent files

PHASE 1: INPUTS (15-30 minutes)
───────────────────────────────
□ Export historical data from HubSpot/GA4/ad platforms
□ Complete campaign brief template
□ Confirm budget and timeline
□ Provide to Orchestrator

PHASE 2: STRATEGY (30-60 minutes)
─────────────────────────────────
Orchestrator → GTM Strategist

□ GTM Strategist analyzes historical data
□ Recommends channel mix with data justification
□ Allocates budget
□ Lists specific deliverables
□ Sets success metrics

Output: GTM Plan document

PHASE 3: PRODUCTION (1-2 hours)
───────────────────────────────
Orchestrator → Production Lead → Sub-agents

□ Production Lead parses GTM Plan
□ Spawns sub-agents in parallel:
  □ Landing page writer → 2 landing pages
  □ Email writer → 5-email sequence
  □ Blog writer → 1 blog post
  □ Social writer → 8 social posts
  □ Ad writer → 10 ad variants
□ Content validator checks all outputs

Output: All content drafts validated

PHASE 4: OPTIMIZATION (30-60 minutes)
─────────────────────────────────────
Orchestrator → SEO Optimizer

□ SEO Optimizer scores all content
□ Identifies critical issues
□ Provides optimized versions
□ Landing pages: SEO 85+, AEO 80+
□ Blog posts: SEO 90+, AEO 85+

Output: Optimized content + scores

PHASE 5: GOVERNANCE (15-30 minutes)
───────────────────────────────────
Orchestrator → Governance Reviewer

□ Brand check: Tone, terminology, messaging
□ Budget check: Allocations within limits
□ Compliance check: Disclosures, legal requirements

Output: Approval or revision requests

PHASE 6: FINAL REVIEW (30-60 minutes)
─────────────────────────────────────
You review the complete package:

□ GTM Plan makes sense
□ Content quality meets standards
□ Optimization scores acceptable
□ Governance checks pass

TOTAL TIME: 4-6 hours

vs Traditional: 2-3 weeks
═══════════════════════════════════════════════════════════════════════════════

The setup is one-time. Once you have the folder structure and skill files, every campaign follows the same workflow. The agents and sub-agents do the production work. You review and approve.

Action item: Run through this workflow with a real campaign. Track actual time spent at each phase. Compare to your previous campaign production time.

Final Takeaways

Skills encode domain knowledge that Claude loads automatically. Write your brand-voice and content-creation skills first. They apply to every campaign.

Agents are specialized assistants for specific phases. The orchestrator coordinates. The GTM strategist plans. The production lead manages content creation. The SEO optimizer polishes. The governance reviewer approves.

Sub-agents handle individual deliverables. The production lead spawns them in parallel. Landing page writer, email writer, blog writer, social writer, ad writer. They all run at the same time.

The folder structure scales with your needs. Start with two skills and one agent. Add more as you validate the workflow. Every campaign you run improves the system.

This structure works in Claude Code, Cowork, and Claude.ai. Technical users get the full automated experience. Non-technical users upload skills and run agents manually. The time savings work either way.