Skill by ara.so — Marketing Skills collection

• Automatic routing based on user intent
• Client memory that persists across sessions
• Self-evaluation of outputs before delivery
• Revision loop that improves the skill over time from logged usage patterns

1. Clone the repository:

git clone https://github.com/bryanengel72/marketing-orchestrator.git
cd marketing-orchestrator

1. Install via Claude Skills interface by uploading the project folder
2. Verify installation by checking for these files:

marketing-orchestrator/
├── SKILL.md                          # Main orchestrator
├── references/
│   ├── revision-log.md               # Session tracking
│   └── client-profiles/              # Client data
└── sub-skills/                       # Six specialist routes
    ├── brand-strategy-icp/SKILL.md
    ├── ad-copy-paid-media/SKILL.md
    ├── content-creation/SKILL.md
    ├── repurpose/SKILL.md
    ├── campaign-mode/SKILL.md
    └── github-deploy/SKILL.md

• Handles intake and routing
• Manages client profile lookup/creation
• Performs self-evaluation before delivery
• Tracks revision loop (every 3rd session)

SKILL.md

• brand-strategy-icp/

  — ICP profiles, positioning, brand voice

• ad-copy-paid-media/

  — Platform ads with character limits

• content-creation/

  — Blogs, social posts, email sequences

• repurpose/

  — Multi-platform content adaptation

• campaign-mode/

  — Full campaign architecture

• github-deploy/

  — Git setup for non-developers

• references/revision-log.md

  — Session log for improvement tracking

• references/client-profiles/

  — Persistent client context

"I need Google Ads copy for a new SaaS product"

1. Checks for existing client profile in

   references/client-profiles/{client-name}.md

2. If new, asks intake questions (company name, industry, target audience, etc.)
3. Routes to

   ad-copy-paid-media

   sub-skill
4. Sub-skill generates output
5. Self-evaluates against rubric (audience fit, clarity, voice, format, goal alignment)
6. Logs session to

   references/revision-log.md

7. Delivers final output with quality check signal

• 2026-04-15: Google Ads campaign (5 ad variants)
• 2026-05-02: Blog post series (productivity tips)

• Avoid technical jargon
• Emphasize ROI and time savings

undefined

"📊 Revision reminder: It's been 3 sessions. Would you like me to analyze 
the revision log and suggest an improvement?"

1. Reads

   references/revision-log.md

2. Identifies most frequent pattern (e.g., "intake questions too long")
3. Proposes specific fix
4. Updates relevant sub-skill after approval
5. Logs change with version bump

• Reduced brand-strategy intake from 8 to 5 questions
• Added auto-skip for returning clients with profiles
• Source: Sessions #12, #14, #16 flagged intake length

"Create an ICP profile and positioning doc for my coaching business"

• Structured ICP with demographics, psychographics, pain points
• Positioning framework (category, audience, problem, solution, differentiation)
• Brand voice guide with tone attributes and examples
• Downloadable Word doc with formatted sections

.docx

• Primary: 125/125 ✅
• Headline: 32/40 ✅
• Description: 29/30 ✅

Platform limits are enforced during generation.

"Repurpose this blog post for LinkedIn, Twitter, and Instagram"

• LinkedIn: Professional narrative (1200-1500 chars), hashtags, engagement hook
• Twitter: Thread format (280 chars/tweet), optimized for retweets
• Instagram: Visual-first caption, story-friendly formatting

"Plan a product launch campaign for a new app feature"

1. Defines campaign architecture (awareness → consideration → conversion)
2. Lists all deliverables in funnel sequence
3. Creates each asset with connecting narrative thread
4. Generates launch checklist with dates and owners

undefined

1. Teaser email (1 week before)
2. LinkedIn announcement post (launch day)
3. Demo video script (launch day)
4. Retargeting ad copy (days 3-7)
5. Follow-up email (1 week after)

• Load email into ESP (Day -1)
• Schedule LinkedIn post (Day 0, 9am)
• Activate retargeting ads (Day 3)

undefined

1. Audience Fit: Does this match the client's target audience profile?
2. Clarity: Is the language clear and jargon-appropriate?
3. Voice Consistency: Does it match the client's brand voice?
4. Platform Format: Does it meet platform-specific requirements (char limits, structure)?
5. Goal Alignment: Does it directly support the stated objective?

✅ Quality check: Passed all rubric categories

1. Create folder:

   sub-skills/new-specialist/

2. Add

   SKILL.md

   with specialist instructions:

undefined

• Question 1
• Question 2

1. Check filename matches client name (kebab-case):

   acme-corp.md

2. Verify file location:

   references/client-profiles/acme-corp.md

3. Confirm markdown syntax is valid

references/revision-log.md

1. Ensure file exists at

   references/revision-log.md

2. Check file permissions (must be writable)
3. Verify orchestrator is logging at end of each session

1. Check user request includes clear trigger words (e.g., "ad copy", "campaign", "repurpose")
2. Review routing logic in main

   SKILL.md

3. If ambiguous, orchestrator should ask clarifying question before routing

"Continue working on the Acme Corp campaign from last week"

1. Loads

   references/client-profiles/acme-corp.md

2. Reads "Past Projects" section for context
3. Confirms campaign details with user
4. Routes to appropriate sub-skill

"Create 5 LinkedIn posts about remote work productivity"

content-creation

1. Generates all 5 posts in single session
2. Ensures variety in hooks and formats
3. Numbers each post for easy reference
4. Offers to add to content calendar

"Make that last ad copy more concise"

1. Retrieves previous output from session context
2. Applies edit instruction
3. Re-runs self-evaluation
4. Delivers revised version with quality check

"Create a brand voice guide, then write 3 ads using that voice"

1. Routes to

   brand-strategy-icp

   first
2. Saves output to client profile under "Brand Voice" section
3. Routes to

   ad-copy-paid-media

   with voice guide context
4. Generates ads conforming to established voice

• Orchestrator structures output for Claude's artifact system
• User can download as

  .docx

  ,

  .md

  , or

  .pdf

• Formatting includes headers, bullet lists, tables